Integrated assessment and energy analysis: Quality assurance in multi-criteria analysis of...
Transcript of Integrated assessment and energy analysis: Quality assurance in multi-criteria analysis of...
DTD 5 ARTICLE IN PRESS
Integrated assessment and energy analysis: Quality assurance
in multi-criteria analysis of sustainability
Mario Giampietroa,*, Kozo Mayumib, Giuseppe Mundac
aIstituto Nazionale di Ricerca per gli Alimenti e la Nutrizione, Via Ardeatina 546, 00178 Rome, ItalybUniversity of Tokushima, Faculty of Integrated Arts and Sciences, Tokushima City 770-8502, Japan
cUniversitat Autonoma de Barcelona, Department of Economics and Economic History, Edifici B, 08193 Bellaterra, Spain
Abstract
Science for sustainability policy requires the handling of multi-dimensional and multi-scale analyses. Integrated
assessment is about generating information relevant for decision-making. This generates a divide between two
scientific paradigms: (1) ‘Post-Normal Science’ acknowledges the unavoidable existence of non-equivalent
perceptions and representations of the reality; legitimate but contrasting perspectives found among social actors;
heavy levels of uncertainty. (2) ‘Normal Science’, believes that it is possible to handle in a rigorous and rational
way these challenges and that therefore it is possible to define in substantive terms ‘the best course of action’ for
society. This paper is written to explain the reasons and the tools developed by scientists working within the Post-
Normal Science paradigm.
q 2005 Published by Elsevier Ltd.
1. Introduction
1.1. The context of this paper
The conclusions of the second Biennial International Workshop Advances in Energy Studies include,
among other recommendations, a call for the scientific community to reframe quantitative analyses
within an ecological economics perspective and to develop more effective tools for decision makers [1].
In relation to this point, one of the technical sessions of that workshop—entitled ‘Energy and
Governance’—was dedicated to the development of policy and management tools for energy analysis to
Energy xx (2005) 1–28
www.elsevier.com/locate/energy0360-5442/$ - see front matter q 2005 Published by Elsevier Ltd.
doi:10.1016/j.energy.2005.03.005
* Corresponding author. Tel.: C39 06 51494439; fax: C39 06 51494550.
E-mail address: [email protected] (M. Giampietro).
M. Giampietro et al. / Energy xx (2005) 1–282
DTD 5 ARTICLE IN PRESS
deal with the need of considering simultaneously different dimensions (economic, ecological, social) in
decision making.
This article is an attempt to blend together the content of two papers which were presented in that
session. The two papers were both dedicated to Integrated Assessment* [terms written in bold and with a
star are defined in the glossary]. The original two papers were focusing on two issues: (i) the technical
problems associated with the handling of integrated packages of indicators referring to different scales
and dimensions of analysis; (ii) an overview of policy challenges and available tools for dealing with the
obvious fact that different stakeholders* will carry different legitimate definitions of what should be
considered as an ‘improvement’ or a ‘worsening’ of an existing situation.
In the rest of this paper, we deal with these two issues and their relevance for energy analysis. In
particular, in this introduction we provide a general discussion of science for governance and
implications for the use of quantitative analyses in this field. We raise several issues related to the
operationalization of basic ideas of Post-Normal Science in relation to science for governance. In
Section 2 we provide an example of the problematic quantification of energy flows when perceived and
represented within a system organized in hierarchical levels on multiple scales (a critical appraisal of the
energetics of human labor). In Section 3 we provide typologies of difficulties faced when attempting a
characterization of the performance of electricity generation in a country on a multi-criteria space*.
Finally, in Section 4, we deal with the obvious but often neglected fact that any process which is
generating scientific inputs used for governance should include an explicit task of ‘quality assurance’.
Such a task is required to guarantee transparency and accountability in relation to the integrity and
competence adopted in the process. This is a crucial requisite to obtain, later on, legitimization and social
acceptance for the consequent process of decision making.
1.2. The epistemological challenge implied by ‘science for governance’
In the year 2000, a group of students in economics operating in France established a web site whose
content was against autism* in academic economics [http://www.paecon.net/]. They were against:
(i) economics’ ‘uncontrolled use’ and treatment of mathematics as ‘an end in itself’, and the resulting in
‘autistic science’; (ii) the repressive domination of neoclassical theory and derivative approaches in the
curriculum; (iii) the dogmatic teaching style, which leaves no place for critical and reflective thought. An
excessive hegemonization of a given scientific paradigm carries the risk of determining a strong
‘normalization’ and ‘lock-in’ of the scientific characterization of any problem structuring in that field.
More or less in the same period the same set of issues popped out in other scientific fields. For
example, several discussions can be found in the field of conservation ecology about the risk of using
excessive formalization in analytical models used to assist decision making [2]. In relation to this topic,
Anderson [3] lists three main related points: (1) quantitative analysis is ‘essentially worthless if it is not
translated into effective policy’ [4]; (2) very complicated models are much more difficult to
communicate and this can imply the loss of important information in the interaction between scientists
and decision makers [5,6]; (3) quantitative analyses must be relevant to decision makers. This requires a
pre-analytical agreement between scientists and decision-makers about an appropriate definition of the
problem structuring [7,8].
Within the field of epistemology* such a discussion is a very old one and has been carried out
in relation to ‘science for governance’ for decades [9]. The debate in this field is about how to
define and guarantee ‘quality’ for science. It should be noted that this discussion deals with one
M. Giampietro et al. / Energy xx (2005) 1–28 3
DTD 5 ARTICLE IN PRESS
of the oldest topic in philosophy. Here we can recall Socrates’ line that ‘scientists are those that
know about their own ignorance’. Using an epistemological jargon we can say that the problem
of science for governance is about how to obtain a semiotic closure (a term introduced by Pattee,
[10]) in science. By semiotic closure* we mean the ability to adopt a problem structuring* which
has two elements:
(i)
it must be relevant in relation to goals and stakes. It has to pass a semantic check in relationto the WHY/WHAT questions. That is, the problem structuring adopted in the analysis must
be declared by some legitimizing source as socially and politically relevant; and
(ii)
it must be useful in relation to the operation of a system of control used to guide action. Ithas to pass a syntactic check in relation to HOW/WHAT questions. That is, the problem
structuring adopted in the analysis must be declared as scientifically useful by recognized
experts and consistent with the codified knowledge preserved by the academic establishment.
The evolution of this debate led first to the definition of Normal Science by Khun [11]: a validated
scientific paradigm implies the fulfillment of a set of given characteristics that are used to judge that a
scientific field has reached semiotic closure. That is, Normal Science can be characterized using a set
of requisites defined in the social discourse, which are able to provide social legitimization to the
scientific endeavour.
Later on Funtowicz and Ravetz introduced the concept of Post-Normal Science to indicate the
existence of fields of applications related to the issue of science for governance in which it is not
possible to reach a substantive agreement about whether or not a given scientific field has achieved the
semiotic closure [12]. That is, it is impossible to obtain an uncontested legitimization of a substantive
problem structuring.
The concept of Post-Normal Science will be discussed again later on in Section 4. For the
moment an important lesson implied by epistemological analysis is that a semiotic closure is
about a successful process. A process is a ‘transient’ in the characterization suggested by Ashby
in his An Introduction to Cybernetics [13]. A process has to be considered as something clearly
distinct from the formalization of that process, which is selected by the analyst. In other words,
the formalization of a process generates an image of that process within a given information
space, and therefore it has nothing to do with the complex series of events to which it refers to
[14–16]. In technical jargon this means that by using a given model scientists can simulate within
a set of variables a causal entailment perceived in the reality—for more on this concept see
Chapter 3 on modeling relation theory in Rosen [14]. Such a simulation can be obtained in
different ways (e.g. by using differential equations or by using transition probabilities). In any
case, it requires substituting the real observed system which is generating patterns and changes in
observable qualities, with an image of the system based on a selection of variables and an
inferential system which is generating patterns of change in the value taken by the proxy
variables [14]. This implies that any formalization requires and reflects a pre-analytical definition
of a finite set of relevant variables which are associated with the identity of the observed system
at a given scale*—e.g. model duration and spatial domain [extent] and differentials adopted in
the equations [grain]; [14,16,17]. Because of this, the resulting formal information space used to
represent the modeled system and its behaviour will reflect not only the characteristics of the
observed system, but also the choices made by the scientists on how to observe reality [18–22].
M. Giampietro et al. / Energy xx (2005) 1–284
DTD 5 ARTICLE IN PRESS
1.3. The difference between ‘Post-Normal Science’ and ‘Normal Science’ approach
The issues discussed so far generated a clear split in the academic communities of scientists using
quantitative analyses in the field of sustainability. The divide is between The Post-Normal Science
approach and the Normal Science approach.
†
the Post-Normal Science approach—There are scientists that notwithstanding their hardscience background have no problems in acknowledging two crucial facts: (i) it is necessary to
deal always with a preliminary semantic check when defining a problem structuring related to
analysis of sustainability. Concepts such as ‘health’, ‘quality of life’, ‘sustainability’ cannot be
defined in a substantive formal way once and for all. Each application requires a special
semantic check within the given context and in relation to the meanings assigned to these
terms by the actual actors; and (ii) quantitative analyses of future scenarios will always be
affected by important doses of ignorance. Nobody can predict the future, no matter how
sophisticated are the models, modelers and computers used to do that. Ignorance about the
future is unavoidable.
†
The Normal Science approach—There are scientists that, on the contrary, believe that it is possibleto define in substantive terms what is good and bad for ‘consumers’, ‘citizens’ and ‘society’ (one
definition fits all). Moreover they assume that: (a) such a definition can be known by them; (b) such
a definition will not change later on; and (c) the issue of the unavoidable presence of uncertainty and
ignorance can be dealt with by using more sophisticated analyses, larger computers, more rigorous
tests and better expertise. This implies assuming that it is possible to deal with the issue of
sustainability in terms of optimization of utility functions, optimization of production functions,
maximization of the efficiency in the use of resources and by engineering ecosystems to improve
their sustainability. Within this paradigm, the quality of the process used to individuate and
implement all these optimizing processes can be guaranteed by the diligent application of scientific
standards based on the ‘state of the art’ of the ‘know how’ available in the different scientific
disciplines involved.
These two approaches generate an important difference.
Scientists working within the paradigm of Normal Science seem to believe that: (i) the
quantitative analysis they are handling represent a substantive definition of what is good and bad
for the system; and (ii) within these quantitative analyses different costs and benefits can be
reduced to each other within a common numeraire (Zthey assume that different indicators
developed in different scientific fields are commensurable). Therefore, they are handling a process
in which the descriptive and normative aspects are melted together. That is, what is calculated to
be ‘the optimum choice’ is supposed to be at the same time: (1) the best possible representation
of the system (this implies the existence of an uncontested agreement among all the actors
involved on the series of choices made on the descriptive side); and (2) the best possible solution
to the problem, which has been chosen among those included in the best possible option space.
This second assumption implies the existence of an uncontested agreement on the normative side
among all the actors involved on: (i) the set of options to consider; (ii) the set of actors to be
considered as relevant; (iii) the set of goals to consider; (iv) the reliability of the information
coming from the descriptive side; (v) the priorities to adopt when making a choice.
M. Giampietro et al. / Energy xx (2005) 1–28 5
DTD 5 ARTICLE IN PRESS
This long list of required agreements among the stakeholders explains why the Normal Science
approach tends to be preferred in decision making. When it is difficult to obtain an agreement among
actors carrying legitimate contrasting views and dealing with issues in which uncertainty is important,
it is much easier to assume that such an agreement does exist rather than verify its existence by asking
them. On the other hand, in the last decades, conflicts over the use of resources and conflicts associated
with the innate tension among different dimensions of sustainability are becoming so important that
they can no longer be ignored. This increasing relevance and visibility of conflicts is forcing scientists
and decision makers to include in their agenda the necessity of dealing explicitly with the evident lack
of agreement expressed by various stakeholders over choices made in the process of decision making
in relation to both the descriptive and normative side. Within this new context, the objective of
scientific investigation must become that of enhancing the process of social resolution of sustainability
issues, rather than individuating a definite technological ‘solution’ or policy implementation.
Moreover, such a process should be based on participation and mutual learning among stakeholders.
Therefore, scientists working within the paradigm of Post-Normal Science have to face two
different typologies of problems. Not only they have to check the quality of their problem structuring
in relation to the special context (‘food quality’ in Burkina Faso is different from ‘food quality’ in the
USA) and in relation to the specificity of the actors (when operating within the USA, ‘food quality’ for
a macrobiotic yuppie is different from ‘food quality’ for a poor single mother of six), but also they
have to perform such a check in relation to two distinct sides of the process of decision making: (1) on
the descriptive side (the WHAT/HOW side)—this implies deciding which is the most useful
representation of the problem—after issue identification. That is, what are the key indicators useful to
characterize the system in relation to the attributes considered as relevant, what are the key
mechanisms and controls associated to the outcomes of interest. This translates into a series of
decisions about the variables to be used, the models to be used, the data and measurement scheme to be
used to characterize and represent possible solutions and scenarios; (2) on the normative side (the
WHY/HOW side)—this implies deciding which is the most relevant issue definition, which is the most
desirable decision to be made within the selected option space. This translates into a series of answers
to questions like: what are the goals and values that should be considered in the choice and
implementation of a given policy? What are the options that should be evaluated? Who are the actors
that have to be invited to be part of the process of discussion? To answer these questions we have to
adopt a series of decisions about how to handle the legitimate but contrasting perspectives found
among the social actors, how to select the targets to be adopted in the process, how to decide the
weighting factors to be adopted when considering contrasting indications and incommensurable trade-
offs. To make the life of Post-Normal scientists harder these two series of decisions to be taken on the
descriptive and normative side: (i) depend on each other (they are linked in a sort of chicken–egg
relation); and (ii) have to be made having clear in mind that the information used in this process is
likely to be affected by high level of uncertainty or even worse by genuine ignorance.
The present authors have been working for quite a while now with quantitative analyses applied to
the issues of sustainability within the Post-Normal Paradigm. As done in the special session of the
Porto Venere workshop, we want to illustrate the existence of possible alternative approaches to the
dominant paradigm of Normal Science in the field of science for governance of sustainability. Our
selection of topics reflects our different background (the first and second authors started their career as
Engineers and then entered into interdisciplinary and trans-disciplinary studies, whereas the third
author started as an Econometrician that later on entered into economic theory, multi-criteria analysis
M. Giampietro et al. / Energy xx (2005) 1–286
DTD 5 ARTICLE IN PRESS
and urban ecology). All of us did a trajectory across disciplines while keeping working on the
generation of practical tools for analysis. This led us to work on the two different sides of the process
of decision making. In particular:
(1)
on the descriptive side—Mario Giampietro and Kozo Mayumi have dealt for many years with thechallenge implied by the use of integrated packages of indicators which are referring to non-
equivalent descriptive domains*. Whenever, an integrated assessment requires the use of
indicators and models that cannot be reduced to each other, we are facing technical
incommensurability*. This means that when comparing different indicators referring to different
typologies of costs and benefits defined at different scales (e.g. improvements and worsening in
my back-yard versus improvements and worsening at the global level) and/or defined within
different scientific disciplines (e.g. economic losses measured in US$ 1998 versus losses of
biodiversity occurring over centuries) it is impossible to develop a system of accounting that is
able to reduce these different typologies of costs and benefits to a common numeraire in a
substantive way.
What if we accept the existence of technical incommensurability in integrated assessment? This is
a class of problems associated with Multi-Scale Integrated Analysis—for an overview of this
concept see Giampietro and Ramos [23]; Giampietro [21]; Giampietro and Mayumi, [24–25].
(2)
On the normative side—Giuseppe Munda has dealt for many years with the challenge implied bythe handling of the following questions. When discussing of sustainability, we should be able, first
of all, to answer questions like: ‘sustainability of what? Sustainability for whom? Sustainability
for how long? Sustainability at what cost?’ [2, p. 26]. This is a series of questions about which it is
impossible to reach a substantive agreement among all relevant actors when coming to the
individuation of a unique vector of answers. Actually, when dealing with the issue of sustainability
and related conflicts it is not even easy to define: (i) who should be considered as a relevant actor,
and (ii) who is legitimized to decide about it. For example, when dealing with the choice, made at
the national level, of energy sources, how to deal with the fact that such a choice may generate
effects on other countries? What about possible effects on future generations? Whenever, an
integrated assessment requires the use of normative values (cultural identities, goals, taboos) that
cannot be reduced to each other we are facing social incommensurability*. This implies that in a
given social conflict we should expect that the ‘terrorists’ on one side are the ‘freedom fighters’ of
the other side. Therefore, when dealing with an existing conflict, it is ludicrous to expect that after
adopting a formal identity for characterizing them in a model (either in the category of terrorist or
in that of freedom-fighter) it will be possible to obtain, later on, an agreement on such a substantive
characterization from the actors operating on the two sides.
What if we accept the existence of social incommensurability in integrated assessment? This is a
class of problems associated with the implementation of procedures aimed at Social Multi-criteria
Evaluation—for an overview of this concept see Munda [26].
An integrated assessment related to sustainability almost always implies the handling of technical
incommensurability and social incommensurability. Scientists willing to use quantitative analysis in
this field have to learn how to better listen to both: (i) other scientists working in other disciplinary
fields (no matter if hard or soft); and (ii) regular people representing the clients of such an assessment.
Acknowledging this fact implies changing the scope of quantitative analysis. Quantitative analyses
M. Giampietro et al. / Energy xx (2005) 1–28 7
DTD 5 ARTICLE IN PRESS
should no longer be used for the individuation of the ‘best course of action’ (a task justified by
substantive reasoning). Rather they should be used for fostering social learning (e.g. looking for better
issue definitions, better understanding of existing trends, individuation of areas of uncertainty, helping
the sharing of meaning among the actors about a useful problem structuring).
1.4. Integrated assessment and engineers
The paradigm of Normal Science so far has been characterized as a scientific paradigm in which
quantitative analyses are used to seek and find the solution that produces more and better than the
actual one. This paradigm, therefore, seems to imply that it is always possible to have: (a) a win-win
solution in the first place (Zmore and better can be obtained without negative effects); and (b) a
substantive formal definition, agreed upon by all social actors, of how to define and measure better. If
these two assumptions were true, the only problem for hard scientists would be that of generating a
formal output (i.e. a number) that indicates the maximum in improvement for the system. Such a final
result is then expected to be an uncontested input for policy-making.
It is fair to say, however, that in reality very few scientists (even those that are proud to call
themselves reductionist) are so fundamentalist to really believe these assumptions (even though very
few of them will write it down). Everybody knows that in reality indications given by models and data
used in any assessment are always mediated by political negotiation and common sense. The real issue
is how to handle this mediation.
Fortunately, among the various categories of hard scientists, engineers are, in general, those better
acquainted with the idea that life is complex. Engineers, like physicians, are a special kind of scientists
that can openly challenge the validity of the two above assumptions. Any good engineer, for example,
knows that optimization in reality means looking for some reasonable compromise and that technical,
legal, political and economic issues are always deeply connected in real situations. As George E. Box
reminds us: ‘all models are wrong, some are useful’ [18]. Indeed, when dealing with science for
governance the real issue is not that of relying on the indications given by models, but rather how to
understand and decide which models can be useful for policy-making, and how to define what should
be considered an acceptable compromise among legitimate but contrasting definitions of
improvements.
Regarding the need of using a multi-criterial approach, it should be noted that in 1824, well
before the introduction of the concept of Integrated Assessment, Carnot stated in the closing
paragraph of his Reflections on the motive power of fire, and on machines fitted to develop that
power: ‘We should not expect ever to utilize in practice all the motive power of combustibles.
The attempts made to attain this result would be far more harmful than useful if they caused
other important considerations to be neglected. The economy of the combustible [efficiency] is
only one of the conditions to be fulfilled in heat-engines. In many cases it is only secondary. It
should often give precedence to safety, to strength, to the durability of the engine, to the small
space, which it must occupy, to small cost of installation, etc. To know how to appreciate in each
case, at their true value, the considerations of convenience and economy which may present
themselves; to know how to discern the more important of those which are only secondary; to
balance them properly against each other; in order to attain the best results by the simplest
means; such should be the leading characteristics of the man called to direct, to co-ordinate the
labours of his fellow men, to make them co-operate towards a useful end, whatsoever it may
M. Giampietro et al. / Energy xx (2005) 1–288
DTD 5 ARTICLE IN PRESS
be (italics added)’ [27, p. 59]. In this conclusive passage of his seminal work Carnot decided to
explicitly warn his readers against the potential pitfalls of analyses that go for a single optimizing
goal (e.g. maximizing efficiency) at the time, without considering the context and the input
coming from the normative side (WHY/WHAT) associated to the social aspects.
Regarding the development and use of analytical tools able to handle several indicators for
characterizing performance in relation to multiple attributes there are rigorous methods for
decision support analysis developed in the field of engineering under the name of Multi-Objective
Decision Making or Multi-Attribute Decision Making (e.g. Scott and Antonsson [28]). It should
be noted, however, that the big advantage of industrial design over analyses of sustainability is
that: (1) all the relevant information for defining the performance of the designed system is
supposed to be available to the designer; and (2) the validity check about the relevance of the
problem structuring (the quality check on the normative side) is supposed to be given.
Finally, one of the most interesting attempt to develop a procedure for dealing with the
challenge discussed so far in participatory way (the use of quantitative analyses within a Post-
Normal Paradigm) has been developed within System Engineering by Checkland and others under
the name of soft systems methodology (Checkland [29], Checkland and Scholes [30], Roling and
Wagemakers [31]).
2. Dealing with the predicament of multiple-scales
2.1. Exploring one of the largest ‘fiasco’ of energy analysis
In order to illustrate the implications of the existence of multiple scales on conventional
energy analysis we are using here an example discussed more in detail in Giampietro and
Mayumi [32]. The example is related to one of the most well-known case studies of this field:
the attempt to develop a standardized tool kit for dealing with the energetics of human labor.
Probably, this attempt represents one of the largest ‘fiasco’ of energy analysis (for an overview of
issues, attempts and critical appraisal of results see: Fluck [33,34]; Giampietro and Pimentel
[35,36]; Giampietro et al. [37,38]).
In order to characterize in useful terms indices of ‘efficiency’ or ‘efficacy’ in relation to the
energetic of human labor three pieces of information are required:
(1)
the requirement—and/or the availability—of an adequate energy input needed to sustain theconversion of interest (an inflow of energy carriers). In the case of human labor this is the flow of
nutrients contained in food.
(2)
the ability of the considered converter to transform the energy input into a flow of usefulenergy to fulfil a given set of tasks (in this case a system made up of humans has to be able
to convert available food energy input into useful energy at a certain rate, depending on the
assigned task).
(3)
the achievement obtained by the work done—the results associated with the application of usefulenergy to a given set of tasks (in this case this has to do with the usefulness of the work done by
human labor in the interaction with the context).
Fig. 1. Relevant qualities for characterizing the behavior of an energy converter across hierarchical levels.
M. Giampietro et al. / Energy xx (2005) 1–28 9
DTD 5 ARTICLE IN PRESS
At this point, if we want to use indices based on energy analysis to formalize the concept of
performance, then we have to link these three pieces of information to four non-equivalent numerical
assessments—Fig. 1:
(1)
Energy input required and consumed by the converter.In the case of human labor, this is food(energy carriers for humans). However, if the converter were a diesel engine, food would no
longer be considered as an energy input. This implies that the energy input can only be
characterized and defined in relation to the chosen/given identity of the converter using it.
(2)
The power level at which useful energy is generated by the converter. This is a more elusive‘observable quality’ of the converter. Still this information is crucial. When dealing with the
characterization of energy converters we have to always consider both the pace of the throughput
(the power) and the output/input ratio. A higher power level tends to be associated to a lower
output/input ratio (e.g. the faster you drive, the lower the mileage of your car). However, it is very
difficult to find a standard definition of ‘power level’ applicable to the generic category of ‘energy
converters’ (e.g. how to compare in terms of power levels: human workers to light bulbs and/or
crop fields to economic sectors?). This is especially true when dealing with multiple converters
operating simultaneously at different scales on multiple different tasks (e.g. ecosystems versus
economies). Moreover, a particular assessment of power level (e.g. 1000 HP of a tractor versus
0.1 HP of a human worker) does not map onto either energy input flows [Zhow much energy is
consumed by the converter over a given period of time used as reference—e.g. a year] nor onto
how much applied power is delivered [Zhow much useful energy has been generated by
M. Giampietro et al. / Energy xx (2005) 1–2810
DTD 5 ARTICLE IN PRESS
a converter over a given period of time used as reference—e.g. a year]. These two different pieces
of information depend on how many hours the tractors or the human worker have worked during
the reference period and how wisely they have been operating.
(3)
The flow of applied power generated by the conversion. The numerical mapping of this qualityclearly depends heavily on the previous choices about how to define and measure power levels and
power supply. In fact, an assessment of the flow of applied power represents a possible
formalization of the semantic concept of ‘useful energy’. Therefore, this is a ‘measured flow of
energy’ generated by a given converter which must be able to fulfil a specified task (e.g. water
pumped out of the well, hectares of tilled soil). To make things more difficult, the definition of
usefulness of such a task can only be given, when considering the hierarchical level of the whole
system to which the converter belong (e.g. if the water is needed to do something by the owner of
the pump, if the hectares are under cultivation). Put in another way, such a task must be defined as
‘useful’ by an observer which is operating at a hierarchical level higher than the level at which the
converter is transforming energy input into useful energy. What is produced by the work of a
tractor has a value, which is defined by the interaction of the farm with a larger context (e.g. the
selling of products on the market). That is, the definition of ‘usefulness’ refers to the interaction of
the whole system—the farm, seen as a black box (to which the converter within the tractor belongs
as a component) with its context. Any assessment of the quality ‘usefulness’ requires a descriptive
domain different from that used to represent the conversion at the level of the converter [21].
(4)
The work done by the flow of applied power (what is achieved by the physical effort generated bythe converter). Work is another very elusive quality that requires a lot of assumptions to be
measured and quantified in biophysical terms. This represents a big problem with energy analysis.
In fact, even if the two assessments #3 and #4 use the same measurement unit (e.g. MJ) they are
different in terms of what are the relevant observable qualities of the system. That is, assessment
#3 (flow of applied power as seen from the converter—e.g. MJ of power delivered by a 100 HP
pump working for 25 h) and assessment #4 (work done—e.g. MJ equivalent to the lifting of a
certain number of m3 of water up to 3 m) neither coincide in numerical terms or map 1 to 1 in a
substantive way. In fact, the same amount of applied power can imply differences in achievement,
because of differences in design of technology (i.e. the pump) and in know how when using it.
2.2. Interpretation of the energy analysis of human labour using hierarchy theory
The overview provided in Fig. 1 should already make clear to those familiar with epistemological
implications of hierarchy theory* (for an overview of the literature see: Allen and Starr [19]; Salthe
[39]; Ahl and Allen [20]) that practical procedures used to generate numerical assessments within a
linear input/output framework cannot escape the unavoidable ambiguity and arbitrariness implied by
the hierarchical nature of complex systems. A linear characterization of input/output using the four
assessments discussed so far requires the simultaneous use of at least two non-equivalent and non-
reducible descriptive domains. This opens the door to technical incommensurability and therefore to
an unavoidable degree of arbitrariness in the problem structuring. Any definition or assessment of
energy flows (both as input and/or output) will in fact depend on an arbitrary choice made by the
analyst about what should be considered as the focal level n. That is: what should be considered as a
converter, what should be considered as an energy carrier, what should be considered as the whole
Table 1
Non-equivalent assessments of energy requirement for 1 h of human labor
Methoda no. Worker system boundar-
iesbEnergy input (EI) per hour of labor
Food energy inputc (MJ/
h)
Exosomatic energy
inputd (MJ/h)
Other energy forms
across scalese
1 Man 0.5f Ignored Ignored
1 Woman 0.3g Ignored Ignored
1 Adult 0.4h Ignored Ignored
2 Man 0.8f Ignored Ignored
2 Woman 0.5g Ignored Ignored
2 Adult 0.6h Ignored Ignored
3 Man 1.6f Ignored Ignored
3 Woman 1.2g Ignored Ignored
3 Adult 1.3h Ignored Ignored
4 Man 2.5f Ignored Ignored
4 Woman 1.8g Ignored Ignored
4 Adult 2.1h Ignored Ignored
5 Household 3.9i Ignored Ignored
6 Society 4.2j Ignored Ignored
7a Household 3.9i 39 (food system)k Ignored
7b Society 4.2j 42 (food system)k Ignored
8 Society Ignored 400 (society)l Ignored
9 Society Ignored 400 (society)l 2!1010 Emjoulesm
After Giampietro [21].a The nine methods considered are: (1) only extra metabolic energy due to the actual work (total energy consumption minus metabolic rate) in
an hour; (2) total metabolic energy spent during actual work (including metabolic rate) in an hour; (3) metabolic energy spent in a typical work
day divided by the hours worked in that day; (4) metabolic energy spent in a year divided by the hours worked in that year. (5) as Method 4 but
applied at hierarchical level of household; (6) as Method 4 but applied at the level of society; (7) besides food energy including also exosomatic
energy spent in food system per year divided by the hours of work in that year (7a at household level; 7b at society level); (8) total exosomatic
energy consumed by society in a year divided by the hours of work delivered in that society in that year; (9) assessing the EMjoules of solar
energy equivalent of the amount of fossil energy assessed with Method 8.b The ‘systems delivering work’ considered are: typical adult man (Man), typical adult woman (Woman), average adult (Adult), typical
household, and an entire society. Definitions of typical are arbitrary and only serve to exemplify methods of calculation.c Food energy input is approximated by the metabolic energy requirement. Given the nature of the diet and food losses during consumption,
this flow can be translated into food energy consumption. Considering also post-harvest food losses and pre-harvest crop-losses, it can then be
translated into different requirements of food (energy) production.d We report here an assessment referring only to fossil energy input.e Other energy forms, acting now or in the past, which are (were) relevant for the current stabilization of the described system even if
operating on space-time scales not detected by the actual definition of identity for the system.f Based on a basal metabolic rate for adult men of BMRZ0.0485W C3.67 MJ/day (WZweightZ70 kg)Z7.065 MJ/dayZ0.294 MJ/h.
Physical activity factor for moderate occupational work (classified as moderate) 2.7!BMR. Average daily physical activity factor 1.78!BMR
(moderate occupational work). Source: Anonymous, 1985. Occupational work load: 8 h/day considering work days only, and 5 h/day average
work load over the entire year including weekends, holidays, absence.g Based on a basal metabolic rate for adult women of BMRZ0.0364 WC3.47 (WZweightZ55 kg)Z5.472 MJ/dayZ0.228 MJ/h. Physical
activity factor for moderate occupational work 2.2!BMR, average daily physical activity factor (based on moderate occupational work) 1.64!BMR. Source: Anonymous, 1985. Occupational work load: 8 h/day (considering work days only) or 5 h/days (average work load over the entire
year including weekends, holidays, absence).h Assuming a 50% gender ratio.i A typical household is arbitrarily assumed to consist of one adult male (70 kg, moderate occupational activity), one adult female (55 kg,
moderate occupational activity), and two children (male of 12 years and female of 9 years).j This assessment refers to the USA. In 1993, the food energy requirement was 910,000 TJ/year and the work supply 215 billion hours.k Assuming 10 MJ of fossil energy spent in the food system per 1 MJ of food consumed (Giampietro et al., 1994).l Assuming a total primary energy supply in 1993 in the USA (including the energy sector) of 85,000,000 TJ divided by a work supply of 215
billion hours.m Assuming a transformity ratio for fossil energy of 50,000,000 EMJoules/Joule.
M. Giampietro et al. / Energy xx (2005) 1–28 11
DTD 5 ARTICLE IN PRESS
M. Giampietro et al. / Energy xx (2005) 1–2812
DTD 5 ARTICLE IN PRESS
system to which the converter belongs, what has to be included and excluded in the characterization of
the environment, when checking: (i) the admissibility of boundary conditions; and (ii) the usefulness
of work done. In terms of hierarchy theory we can describe this fact as follows.
The unavoidable preliminary triadic filtering* needed to obtain a meaningful representation of
reality implies selecting: (i) the interface between the focal level n and the lower level nK1 to
represent the structural organization of the system (WHAT/HOW); and (ii) the interface between
the focal level n and higher level nC1 to represent the relational functions of the system
(WHAT/WHY). The arbitrariness of this choice is at the basis of the impasse faced by energy
analysis. Ignoring this fact, simply leads to a list of non-equivalent and non-reducible
assessments of the same concepts. A self-explanatory example of this standard impasse in the
field of energy analysis is given in Table 1—which reports an example of several non-equivalent
rigorous assessments of the energetic equivalent of 1 h of labor found in literature.
That is, every time we choose a particular hierarchical level of analysis for assessing an
energy flow (e.g. an individual worker over a day) we are also selecting a space-time scale at
which we will describe the process of energy conversion. This, in turn, implies a non-equivalent
definition of what is the context (environment) and what is the lower level (where structural
components are defined). Human workers can be seen as individuals operating over a 1 h time
horizon (muscles are the converters in this case), or as citizens of developed countries (machines
are the converters in this case). The definitions of identities of these elements must be, then
compatible with the identity of energy carriers (individuals eat food, developed countries eat
fossil energy).
A discussion of an innovative approach to handle technical incommensurability when dealing with
multi-scale integrated analysis of metabolic systems is available in three chapters (Chapters, 6, 7 and 8
co-authored with Kozo Mayumi) of the book of Giampietro [21]. These chapters present an analysis of
such a problem and possible ways out (e.g. looking for mosaic effects across levels, impredicative loop
analysis, narratives useful for surfing in complex time). A brief overview of these concepts is now
available in Giampietro and Ramos-Martin [23].
3. Facing the predicament of multi-criteria energy analysis
3.1. The unavoidable presence of both technical and social incommensurability
In the previous section, we discussed the problem associated with the insurgence of technical
incommensurability, even when adopting a single criterion of analysis (energetics of human labor),
when multiple scales are considered. In this section, we want to discuss the fact that as soon as energy
analysis deals with problems that are multi-dimensional and multi-scale (the quote of Carnot can be
recalled here) it is unavoidable to face not only technical but also social incommensurability.
To introduce an example of multi-criteria energy analysis we will use a typical ‘problem’ that can
only be characterized in relation to non-equivalent attributes. The example is: ‘assessing the
performance of electric generation within a society’ in order to decide about technical innovations or
new regulations. To characterize the performance of the energy sector one has to consider a set of non-
reducible criteria, which are related to non-equivalent objectives (Multi-Dimensional Analysis). In
our example we may select: (i) self-reliance for the country; (ii) low economic cost for the supply of
M. Giampietro et al. / Energy xx (2005) 1–28 13
DTD 5 ARTICLE IN PRESS
energy carriers; (iii) good health of ecosystems; (iv) high quality of life for citizens; (v) minimization
of risk of accidents; (vi) preservation of landscapes. On the descriptive side, an integrated analysis of
potential changes in the energy sector then requires gathering different indicators reflecting changes in
relation to these criteria. This requires an interdisciplinary team of scientists. Several distinct
disciplines, in fact, are required to study in quantitative terms (with numerical indicators and models)
the expected effects of a given policy in relation to the given set of relevant attributes. Obviously, the
more we enlarge the set of dimensions and scales to be included in the characterization and assessment
of scenarios of change in the energy sector, the more we make the task of scientists harder. The more
the effects of technical and social incommensurability enhance each other, the more it becomes
difficult to guarantee the quality of the process of Integrated Assessment. On the other hand, concepts
like welfare and sustainability are multi-dimensional concepts in their very essence.
3.2. An example of multi-criteria evaluation of the electric sector of a country
Let’s imagine for example to discuss of a research protocol for a multi-criteria analysis of different
options for generating electricity. This would require:
(i)
Using parallel descriptions of processes referring to the various distinct hierarchical levels atwhich the various energy conversions can be analysed (e.g. individual piece of machinery, the
whole plant, the level of the related economic activity, the macro-economic level, and the
biosphere).
(ii)
Using several sets of indicators of ‘good’ and ‘bad’ referring to distinct hierarchical levels andselected in order to include relevant perspectives of various actors (e.g. owner of the plant,
investors, workers, consumers, future generations, national governments, local communities,
marginal social groups, ecosystems, endangered species).
(iii)
Establishing links, in the integrated analysis, among processes occurring on the varioushierarchical levels considered in the analysis. This is a step required to enable an informed
discussion on possible future scenarios following technological and policy changes (e.g. what
happens when pulling in different directions a blanket which is too short).
(iv)
Involving stakeholders in the process of definition of the terms of reference for the analysis, in theselection of indicators, in a critical appraisal of data and models generating scenarios and, finally,
in a discussion of pros and cons of the various options considered.
Remaining in the example of a multi-criteria analysis of the performance of energy sectors, let’s
now imagine that one wishes to either: (A) evaluate ex-post three strategies adopted in different
countries for producing electricity in the second half of 1900—e.g. France, Italy and Spain; or
(B) evaluate now pros and cons of three possible options for producing electricity—e.g. using the
same three options chosen in the past by these countries: (i) nuclear; (ii) oilC import; (iii) use as much
as possible coal. In order to do such an evaluation we should, first of all, select a set of indicators able
to cover the various relevant dimensions required to characterize the performance of a system
producing electricity. Then we should be able to: (a) assign numerical values to the resulting set of
indicators; and (b) judge the reliability of these assessments.
For the sake of simplicity, we select in our hypothetical analysis four main objectives for the
evaluation of electricity production: (1) economic performance; (2) environmental impact; (3) political
Economic characteristics Ecological Impact
Require
ment
of Subsid
ies
Return on
InvestmentGHGemission
Cos
tU
S$/k
Wh
(giv
en y
ear)
Radioactive
wastes
Habitat
destruction
Pro
babi
lity
of a
n ac
cide
nt
Dependency
on imports
Central controlon the supply
Aestheti
c
effec
ts on
landsca
pe
Consequences
of an accident
Internal supply
of plutonium
France
Nuclear
Italy
Oil/Import
Spain
Coal
Safety/Quality of lifeSelf-reliance/Power
Fig. 2. Multi-objective integrated representation of the performance of an electric sector.
M. Giampietro et al. / Energy xx (2005) 1–2814
DTD 5 ARTICLE IN PRESS
relevance of self-reliance and military strength; (4) safety and quality of life for consumers. After
having done that, a characterization on such a multi-objective space would look like what is shown in
Fig. 2.
We apologize in advance for the over-simplification in the characterization of the options given in
Fig. 2, but getting into details or examples of real studies is beside the point here. As it will become
immediately evident from the following discussion, the specific quality of such an initial
characterization is not particularly relevant. Its role is just that of providing a narrative to our
discussion. Our main point is exactly that it is not possible to do an integrated analysis of the type
shown in Fig. 2 in ‘the right way’ (this would imply operating within the paradigm of substantive
rationality). No matter how good is the protocol specified for such an analysis, it is unavoidable that,
according to the perspective, data and personal opinions of some other analyst, such a characterization
could have been done in a better way. In fact, the chain of choices that leads to the representation given
in Fig. 2 is necessarily associated with a chain of location specific events and with a series of value
calls. These value calls start with the selection of the four objectives.
Probably, each informed reader of Energy if asked to do so, would have proposed a different
selection of objectives and/or indicators for characterising the performance of electricity generation.
But each one of these different characterizations of an energy sector based on the approach given in
Fig. 2 could have been easily criticised by other readers. The point we want to make with this example
is exactly that any characterization of the performance of an energy sector on a multi-criteria space
will always face the following systemic problems:
M. Giampietro et al. / Energy xx (2005) 1–28 15
DTD 5 ARTICLE IN PRESS
†
Unavoidable ‘openness’ of the information space. Actually, the number of objectives (and relevantcriteria for each objective) that could be used to characterize the performance of electricity
generation is virtually infinite. Depending on the variety of locations in space, locations in time and
the characteristics of the cultural contexts in which the analysis can be performed. This means that
there is an open and expanding universe of possible criteria (and relative indicators), which can be
and are actually used by human actors—and educated readers—to define such a performance.
†
Incommensurability of trade-offs. Some of the criteria (and relative indicators) measuring relevantcharacteristics of the system will result incommensurable (price of a kWh in dollars, degree of self-
reliance, aesthetic preferences, negative impact on biodiversity) and conflicting in nature (e.g.
within a conversion process at a given level of technology, the lower the economic cost for
production, the higher the environmental impact).
†
Indeterminacy. Not all the assessments referring to different indicators have the same reliability(e.g. assessment of the emission of CO2 per Kwh versus assessment of the effect on biodiversity).
The various indicators of performance can imply the adoption of different time horizons for
representing changes in the energy sector. That is, when considering an economic criterion such as
the cost of electricity, we have to consider fluctuations in prices with a ‘time differential’ of 1 year.
Whereas when considering changes in the level of ecological impact we can deal with gradients
with a life-expectancy of centuries or even more—e.g. the expected decay time of radioactive
wastes. When one deals with long-term scenarios about activities that have never been done before
(e.g. the handling of radioactive waste during the decommissioning of nuclear plant) it becomes
very difficult to get reliable indications on the actualised cost for the year 2020—expressed in US
dollars 1998—for 1 kWh of electricity produced now.
†
Genuine ignorance. The existence of various dynamics operating in parallel on different scales is atthe root of uncertainty in both forms: indeterminacy (impossibility to perform accurate
predictions—butterfly effect) and genuine ignorance (possible emergence of new relevant issues
to be dealt with in the future).
†
Quality of the problem structuring (on the normative side). This reflects the agreement from allstakeholders on what is the right ‘problem structuring’ (issue identification, selection of indicators,
quality of data) to be adopted in the analysis. For example, when dealing with the objective
‘self-reliance and military power’ (lower-left quadrant in Fig. 2) we assumed in this radar-diagram
two relevant criteria: (1) ‘internal supply of plutonium’ (which can be used for making nuclear
weapons); and (2) ‘central control on the supply’. In this representation (based on the Flag Model)
these criteria have been considered to be associated with ‘improvements’ when the value taken by
the indicator increases. In Fig. 2 this is reflected by a position taken by the value on the indicator in
the dark grey area close to the origin (red in the original diagramZbad). Whereas a position in the
very light grey away from the origin (green in the original diagramZgood). An overview on the flag
model and more in general on graphic tools for Multi-Objective Integrated Representation is given
in Gomiero and Giampietro [40]. Many people may agree with this choice of defining improvement
in relation to this criterion (e.g. this was the case of decision makers in France in the 1960s). For
these people having the capability of making nuclear weapons and a strong control on the supply of
power is ‘good’. Many others, however, would resent this choice. According to those that oppose
nuclear energy and the associated possibility to make nuclear weapons the characterization of the
nuclear option for generating electricity should adopt an opposite graphic representation of these
M. Giampietro et al. / Energy xx (2005) 1–2816
DTD 5 ARTICLE IN PRESS
gradients (inverting the position of the areas indicating ‘good’ and ‘bad’). In this view, nuclear
capability and central control should be identified as bad characteristics of this option. Obviously,
dealing with such a contrast of opinions is well outside the realm of scientific analysis. Still, such a
decision is required by the analyst at the moment of generating an integrated representation of
performance as done in Fig. 2.
†
Quality of data (on the descriptive side). As noted earlier the set of assessments used in the variousquadrants are subject to different doses of arbitrariness. But even remaining within the same
quadrant and within the same indicator, how accurate is the assessment—for example—of the
economic cost of a kWh of electricity? Any numerical assessment of such a cost will depend on
what the analyst decide to include in the relative calculation (e.g. the time horizon and the discount
rate adopted and the procedure adopted for Life Cycle Assessment, the risk of environmental
damages). This is the analogous to the truncation problem in energy analysis (what is accounted as
embodied in the assessment of a given energy input). A paper of Andrew Stirling, focusing exactly
on the range of values found in the assessment of economic costs for producing a KWh of
electricity, is enlightening at this regard. Stirling [41] monitored a large number of actual studies,
sponsored by industries or governments in industrialized countries. They were all aiming exactly at
the assessment of ‘the economic cost’ of electricity generation (the same indicator referring to the
same criterion). When assessing external costs (e.g. related to environmental impact) of electricity
generation in constant US currency terms (US$ 1995) in a new coal power station, the range of
values that Stirling found in literature goes from: ‘less than 0.05 cents/kWh’ to ‘more than
1000 cents/kWh’! The spread of the values taken by these assessments is so large to show a
difference of more than 20,000 times. Stirling’s analysis is very important for our discussion since
each of the assessments considered in his study was performed by reputable scholars operating in
reputable institutions (normal science at work for individual risk assessment studies). These studies
were all providing highly precise estimates based on the use of three or four digits! The resulting
estimates were then used to support strongly prescriptive conclusions (normative input) about how
to select the ‘best option’ considered in the analysis. The mechanism justifying the existence of
large differences in non-equivalent rigorous assessments has been discussed in Table 1.
†
Quality of the process of decision making (on the normative side). Whenever one is in the unlikelysituation of having reached a general agreement on: (a) a satisfying selection of objectives; (b) a
satisfying selection of indicators; (c) the usefulness of the underlying problem structuring; and
(d) the reliability of the set of data used in the representation of performance on the multi-criteria
space; the story is still not over. Then one has to start the process required for deciding what should
be selected as the best profile of different types of ‘costs’ and ‘benefits’ within a set of alternatives
for a particular country when producing electricity. This second process does not coincide with
scientific analysis. For example, the development of nuclear energy in France or the reliance on coal
plants in Spain even if fuelled by low quality local coal, can be explained by a clear priority given to
the objective of self-reliance over the others. The two generals in power at the moment the decision
was made—De Gaulle and Franco—probably did play a key role in determining this priority. On
the contrary, for historic and political reasons, such an objective was never a key issue in Italy.
†
Quality of the handling of uncertainty throughout the whole process. Finally, any selection of anoption based on the discussion of future scenarios is unavoidably affected by heavy doses of
uncertainty. When choosing between the three options represented in Fig. 2, how to deal with the
M. Giampietro et al. / Energy xx (2005) 1–28 17
DTD 5 ARTICLE IN PRESS
different degree of uncertainty affecting the characterizations of these three options associated with
the chosen multi-criteria space? What if some crucial criterion is missing for the moment?
Whenever it is impossible to establish exactly the future state of the problem faced, one can decide
to deal with such a problem either in terms of stochastic uncertainty, thoroughly studied in
probability theory and statistics, or in terms of fuzzy uncertainty, focusing on the ambiguity of the
description of the event itself [42]. However, one should always be aware that genuine ignorance is
always there too. One cannot calculate the risk of ‘something’ happening in the future, without
knowing ahead what this ‘something’ will be.
In all cases in which we can expect that the information used in the problem structuring is affected
by subjectivity, incompleteness and imprecision, the great advantage of multi-criteria evaluation is the
possibility to take these different factors into account.
4. Quality assurance in integrated assessment of sustainability
4.1. The task of quality assurance implied by Post-Normal Science
Let’s imagine, now, to deal with a controversial topic such as: ‘it is nuclear energy a desirable
alternative to oil?’ Let’s imagine, to make things more difficult, to frame such a choice within the
rationale provided by the precautionary principle (which has recently been re-stated as a key guiding
concept for policy in Europe in a Communication from the European Commission [43]). In this case,
we should be able to define, first of all, what should be considered as ‘enough’ scientific evidence
about possible future hazards linked to this option. The crucial dilemma about ‘how to apply’ such a
principle therefore is related to a previous choice of paradigm between substantive rationality and
procedural rationality [44]. Substantive rationality means assuming that it is possible to define in
absolute terms what should be considered ‘enough’ scientific evidence. Whereas, procedural
rationality means acknowledging that it is not possible to define in absolute terms what should be
considered as ‘enough’ scientific evidence. Therefore, such a decision must be ‘the outcome of
appropriate deliberation’ and therefore ‘procedural rationality depends on the process that generated
it’ [44, p. 131]. At this regard it should be noted that in some occasions we will never know, not even
‘ex post’ what course of action would have been to be considered as ‘the best one’. In fact, in real life,
evolving systems imply a long list of ‘only one experiment’.
The paradigm of substantive rationality (the hidden choice of conventional reductionism) implies
that a committee of experts can decide about what should be considered ‘enough’ and therefore the
best interest of the citizen. But, what if the perception of ‘best interest of citizen’ adopted by the
‘committee of experts’ does not coincide with the set of criteria considered relevant by the citizens
themselves? What if the assessment of ‘better efficiency and negligible risk’ provided by the
‘committee of experts’ will turnout to be wrong? [45].
According to what discussed so far, therefore, a Post-Normal Approach to the analysis of the
performance of an electric sector should:
1.
Keep separated the descriptive from the normative side. Scientists should not claim to provide ‘thecorrect’ analysis/description of the system. Rather, what scientists can do is simply to generate
M. Giampietro et al. / Energy xx (2005) 1–2818
DTD 5 ARTICLE IN PRESS
several sets of ‘view dependent’ representations of an energy system (reflecting the interests of
stakeholders) that can be used to discuss pros and cons associated to possible policies. The novelty
of this approach is that when generating these various models scientists should make clear such a
view-dependence from the beginning;
2.
Generate analyses that can learn in time. Any analysis should be organized in order to remain opento additional alternative ‘view-dependent’ representations. In fact, for enhancing the ongoing
negotiation among groups expressing different views and interests about the performance of the
same energy system, it can become useful during the process to add alternative ‘view-dependent’
representations (new variables, indicators, models, sources of data) to the original set.
3.
Acknowledge the unavoidable presence of ignorance and uncertainty. The goals related to theconcept of sustainability can not be all achieved at the same time, just by adopting a single ‘silver
bullet’ technical solution. Rather dealing with the issue of sustainability, implies a wise
compromise between contrasting goals. The goals of flexibility and adaptability very often clash
with the goals of efficiency and economies of scale. This type of dialectics is associated with a large
degree of uncertainty;
4.
Do not put all the epistemological eggs in the same basket. Look as much as possible for mosaics ofcomplementing information (epistemological plurality), using analyses which have been generated
in different scientific fields (physics, engineering, economics, sociology, political science, applied
and theoretical ecology, etc.). If we admit that it is not possible to compress scientific descriptions
of the same system obtained adopting different space-time scales (technical incommensurability),
then we have to look for mix of complementing non-equivalent views. The reader can recall, here,
the integrated use of medical tests (blood tests, X-rays, ultrasound scan, etc.) to get a better picture
of the health of a given person;
5.
Avoid a dramatic hegemonization in the choice of relevant objectives and criteria. In normativeterms, this can be considered an analogous of the advice #4 (avoiding to put all the eggs in the same
basket in the descriptive side). In any process of decision-making, it is always useful to enlarge the
number of alternative perspectives that can be used to define tasks, indicators, models and their
relative relevance. In fact, the issue of sustainability implies considering in the process of problem
structuring the view of actors that in the past were not included in ‘traditional’ cost-benefit
analyses. Therefore, within the Post-Normal Science paradigm the tools developed by neo-classical
economics for dealing with sustainability are considered an example of excessive hegemonization
in normative terms. That is, the interests of future generations, the health of local ecosystems
(which is associated with the preservation of biodiversity), and the preservation of values
associated to non-dominant cultures tend to be systematically neglected by standard economic
accounting.
6.
Increase the transparency of the process of integrated assessment by making scientific analysesmore ‘stakeholder-friendly’. Scientists should try to make an extra effort to make their
assumptions, analyses, data collection and measurement processes easy to understand.
4.2. Procedures and tool kits to be developed
After acknowledging that it is not possible to define in substantive terms the ‘right problem
structuring’ for an Integrated Assessment of a sustainability problem, we are left with the only viable
M. Giampietro et al. / Energy xx (2005) 1–28 19
DTD 5 ARTICLE IN PRESS
option of a procedural definition of it. This opens the problem of how to guarantee the quality of such a
procedural definition. The solution suggested here is an iterative use of two different tool kits for
performing a quality check both on the descriptive and the normative side. In particular we can
imagine a process of iteration among two distinct activities:
(A)
Discussion and development of a tool kit for ‘discussion support’. In this activity scientists arethe main actors and social actors the consultants: the goal is the development of integrated
packages of analytical tools required to do a good job on the descriptive side. This information
space has to be constructed according to the EXTERNAL input received from the social actors
about what is relevant in relation to the definition of good and bad. The social actors, as
consultants, have to provide a package of questions to be answered. In this activity, the scientists
are those in charge to process such an input according to the best available knowledge of the
issue.
This side of the process requires an ability of scientists coming from different disciplines to
interact on a given problem structuring provided by the society. This is what we introduced before
under the label of multiple scale integrated analysis.
(B)
Discussion and development of a tool kit for ‘decision support’. In this activity stakeholders andother relevant agents are the main actors and scientists are the consultants: the goal is the
development of procedures required to do a good job on the normative side. The resulting process
should make possible to decide, through negotiations: (1) what is relevant and what should be
considered as good and bad in the decision process; (2) what is an acceptable quality in the
process generating the information produced by the scientists (e.g. definition of quality criteria:
relevance, fairness in respecting legitimate contrasting views, no cheating with the collection of
data or choice of models), and finally (3) the final outcome of the integrated assessment (e.g.
policy to be implemented).
This side of the process requires an EXTERNAL input (given by scientists) consisting in a
qualitative and quantitative evaluation of the situation on different scales and dimensions. In their
input, scientists have to include also information about expected effects of changes induced by the
decision under analysis (discussion of scenarios and reliability of them). But the social actors are
those in charge to process such an input. This is what we introduced before as social multi-criteria
evaluation.
Since the scientific process associated with activity A affects the social process associated with
activity B and vice versa, the only reasonable option to handle this chicken–egg situation is to
establish some form of organized iteration between the two, by keeping in mind that process A is a
scientific activity referring to the descriptive side (that requires an input from social actors) and
process B is a social activity referring to the normative side (that requires an input from scientists).
Both of them depend on each other. This is where the need of a new type of ‘expertise’ enters into play.
In order to assure the quality of such an iterative process, it is necessary to implement adequate
procedures. The implementation of these procedures requires developing an expertise in relation to
three tasks of quality assurance: (a) the process on the descriptive side determining the quality of
Multi-Scale Integrated Analysis; (b) the process on the normative side determining the quality of the
Societal Multi-Criteria Evaluation; and (c) the handling of iteration among these two processes (about
the fairness and competence used to handle the interaction among the actors in the two processes). For
M. Giampietro et al. / Energy xx (2005) 1–2820
DTD 5 ARTICLE IN PRESS
an example of how it is possible to integrate soft system methodologies (e.g. the approach developed
by Checkland) with quantitative analyses in an iterative process see Chapter 5 of Giampietro [21].
4.3. The limitations of formal multi-criteria characterization
A typical multi-criteria problem (with a discrete number of alternatives) may be described in the
following way: A is a finite set of n feasible actions (or alternatives); m is the number of evaluation
criteria which are considered relevant in a decision problem. In this way a decision problem may be
represented in a tabular or matrix form. The performance of any given alternative, according to the set
of relevant criteria can be characterized through a multi-criteria impact profile in a matrix form, which
is an alphanumeric view of the information organized in Fig. 2.
These multi-criteria impact profiles can be based on quantitative, qualitative or both types of
information. In the example of multi-criteria impact profile given in Fig. 2 we have nZ3 (possible
choices of electricity generation) and mZ12 (criteria used to evaluate the performance), indicated by
the 12 axes on the radar diagram: e.g. cost of kWh, internal supply of plutonium, Green House Gas
emission (GHG), etc. These criteria are grouped into four main dimensions (economic analysis,
ecological impact, self-reliance/power, and safety/quality of life). That is, the concept of dimensions is
hierarchically higher than the concept of criteria. Different objectives may require the adoption of non-
equivalent descriptive domains (different dimensions of analysis). This implies that several criteria
can be used to characterize one dimension (e.g. the economic evaluation of a power plant requires the
assessment of several economic indices). Goals reflect the need of formulating the general objectives
(selected in the profile). This is obtained by locating target values over the set of selected indicators. In
the graphic representation given in Fig. 2, goals are indicated by the various bullets on the various
indicators in the radar diagram (in the middle of the mid-grey area). In this way, we are bridging two
hierarchical levels of analysis: (1) definition of performance in general terms, obtained by selecting a
set of different relevant dimensions (i.e. formulation of general objectives such as maximizing the
economic efficiency, maximizing the usefulness for users, etc.); (2) translation of these general
principles into a numerical mapping of performance over a set of indicators, which are necessarily
context specific (location specific description). In the example of Fig. 2, we have that after indicating
the goals and the scores of the alternative for each criterion, the 12 criteria are transformed into
attributes of performance. They reflect, in this way, the level of achievements of the goals. Finally, we
can imagine that special threshold values (e.g. a limited budget of money for producing electricity, a
given constraint on local availability of natural resources, a strong will of the people about the level of
hazard they are willing to accept, given regulations about maximum level of pollutant acceptable) will
imply the existence of constraints on the value that can be taken by the selected set of indicators. In
this way it is possible to define the viability domain for alternatives. In the example given in Fig. 2, the
viability domain (or feasibility region) would be the area on the radar diagram obtained by excluding
the white circle on the origin of the axis (Ztoo bad to be acceptable).
After having framed a multi-criteria analysis in this way, it is possible to adopt various methods
related to the ranking of alternatives. In relation to this task there are two definitions that carry a lot of
epistemological implications [42]:
(1)
Dominance: an action a dominates an action b if a is at least as good as b for all the criteria takeninto consideration, and much better than b for at least one criterion.
Criterion 1
Criterion 2
A
B
C
E
D
A, B, C, D, E, represent different alternatives
Fig. 3. Example of ranking of solutions.
M. Giampietro et al. / Energy xx (2005) 1–28 21
DTD 5 ARTICLE IN PRESS
(2)
Efficient solution: an action a belonging to A is efficient if there is no action b in A whichdominates a.
Therefore, the concept of multi-criterial efficiency (a formal definition of ‘better’) can easily be
illustrated graphically—see Fig. 3 which refers to a two-criteria state space. Alternative C performs
better than B in all respects and hence C should be preferred to B. The same can be said for D
compared with A. Thus only C and D are efficient, rational, alternatives, compared with B.
The implications of these definitions can be critically appraised against the basic epistemological
problems discussed in the previous three sections. As observed in Section 1, the concepts of
‘dominance’ and ‘efficiency’, which can be used to rank different options, apply not to the reality, but
rather, to a given representation of the reality, determined by the choice made by the analyst when
selecting a given problem structuring. That is, what is ‘seen’ as the existing set of constraints and
opportunities in the formalization of the problem structuring used in the multi-criteria characteriz-
ation, does not necessarily reflect the real set of constraints and opportunities. This distinction between
‘reality’ and ‘representation of the reality’ becomes crucial in all cases in which uncertainty
and genuine ignorance can be assumed to play an important role in the process. This is always the case
when dealing with future scenarios, reflexive systems and multiple relevant scales. In this situation the
analysts should be very aware of the heavy implications of confusing the reality with their
representation of it. This means that the concept of ‘efficient solution’ may be useful, but only in the
short term, whereas it is very likely to become dangerous in the long term. For a detailed analysis of
the mechanism generating Jevon’s paradox*—see Giampietro [21].
Any algorithmic definition of ‘an ideal option’ (an option dominating all the others on all the
selected criteria) must be based on a static and finite information space. This implies the unavoidable
existence of: (1) other goals/objectives (reflecting the identity of other social groups) not considered in
the existing analysis; and (2) other relevant dynamics and constraints which could have been detected
only by adopting a different scale and a different set of observable qualities in the observation space.
This is the reason why it is important to look for several ways for structuring the problem
complementing each other (plurality of representations). Whenever we describe a system, even when
M. Giampietro et al. / Energy xx (2005) 1–2822
DTD 5 ARTICLE IN PRESS
adopting a multi-dimensional representation, we are using a number of dimensions smaller than the
one required to catch all its relevant qualities. The metaphor of the Flatlanders [46] can be a good
example of this. This is what explains the danger of going for the ‘best solution’ as determined by a
computation to define a political decision. The difference from computation and decision making
relies exactly in the fact that decision making is based on both: (1) context dependent goals and
(2) insufficient information (Fesce, personal communication). In this frame we recall here the
suggestion of Simon that when dealing with complexity the concept of ‘procedural rationality’ should
replace that of ‘substantive one’ (the one adopted by default by reductionism). The amplification of the
best performing activities (according to a selected set of goals and a given problem structuring) and the
elimination of less performing activities (considered as obsolete at a given point in space and time), is
a wise strategy only under the assumptions that adopted (1) goals and (2) problem structuring will
remain relevant and useful in the future and wherever such a strategy is applied. As soon as either of
these two assumptions will lose validity the strategy will backfire.
4.4. Addressing explicitly the unavoidable existence of conflicts
In the previous section we saw that technical and social incommensurability associated with large
dose of uncertainty imply a serious limitation on the use of algorithms and formal protocols for
handling in an appropriate way the conflicts on the normative side.
Therefore, given the importance of value conflicts for the class of problems we are dealing with, it
becomes essential also to develop analytical tools showing clearly the impacts of different potential
choices on each of the different stakeholders (or social actors) considered in the analysis. This means
that together with an integrated representation of changes on different levels and different descriptive
domains (something similar to what represented in Fig. 2), we have also to generate a conflict analysis
procedure, able to indicate groups whose interests seem to cluster together or diverge.
An example of analytical tool developed to handle this task is the Novel Approach to Imprecise
Assessment and Decision Environments (NAIADE), which was created for sustainability policy
problems specifically [42]. NAIADE is a discrete multi-criteria method whose impact (or evaluation)
matrix may include either crisp, stochastic or fuzzy measurements of the performance of an alternative
with respect to an evaluation criterion, thus it is very flexible for real-world applications.
A peculiarity of NAIADE is the use of conflict analysis procedures to be integrated with the multi-
criteria results. This to allow policy-makers to seek for decisions that could reduce the degree of
conflict (in order to reach a certain degree of consensus) or that could have a higher degree of equity on
different income groups. NAIADE uses a fuzzy conflict analysis procedure. Starting with a matrix
showing the impacts of different courses of action on each different interest group we can build a
different type of impact matrix, which can be called a social impact matrix. A fuzzy clustering
procedure indicating the groups whose interests are closer in comparison with the other ones can be
used. In this way, by applying the NAIADE software, one can arrive to a dendogram of a coalition
formation process (related to the same problem considered in the multi-criteria impact matrix
referring to the information graphed in Fig. 2). An example of such an analysis (but applied to a
participatory integrated assessment of policy options about water management) is illustrated in Fig. 4.
This different perspective of analysis, obviously, carries the risk of generating serious divergence
between the ranking obtained when using a multi-criteria selection process based on the information
provided by a multi-criteria impact matrix (the type of information given Fig. 2) and the equity
Fig. 4. Example of conflict analysis.
M. Giampietro et al. / Energy xx (2005) 1–28 23
DTD 5 ARTICLE IN PRESS
ranking based on social conflicts represented in the social impact matrix (the type of information given
in Fig. 4).
Multi-criteria impact matrices (characterizing different options in relation to the score of different
attributes on a multi-criteria space) tend to be considered more ‘technical’. In fact, the structure of this
information reflects the choices made in the step represent by scientists that selected relevant criteria,
useful indicators and reliable data. However, as observed before, the ‘neutral’ role of this technical
information is far from being above confrontations. For example, getting back to the integrated
analysis given in Fig. 2, interests group (e.g. a lobby in favour of nuclear energy) can fight about the
relevance of the various indicators or about the reliability of the relative data (the reader can recall
here the amazing findings of Stirling). Moreover, ranking various options requires considering all
the criteria simultaneously in search of the most satisfying compromise solution. So at times the fight
for the protection of individual interests can move to a deliberate effort of including or excluding those
objectives or criteria that are viewed as favourable or dangerous for certain interests.
On the contrary, social impact matrices based on the impact score of each of the considered
alternatives in relation to each interest group are much more direct. In fact, such a score is determined
by the group itself. Irreducible conflicts may exist between different coalitions or even between single
groups. This policy analysis can be conditioned by heavy value judgments such as: have all actors the
same importance (i.e. weight)? Should a social desirable ranking be obtained on the grounds of the
majority principle? Should some veto power be conceded to minorities? Are income distribution
effects important? And so on. That is, this combined analysis makes it possible to address in a more
structured way, some of the topics addressed in the list of quality checks presented before. Examples
of applications are available in the literature [47–51].
M. Giampietro et al. / Energy xx (2005) 1–2824
DTD 5 ARTICLE IN PRESS
4.5. Back to the basic implications of Post-Normal Science
In general terms, we can say that the epistemological concerns discussed so far have not been
considered very relevant by scientific research in the past. On the other hand, the new nature of the
problems faced in this third millennium (e.g. climate change, mad cow and avian flu, genetic modified
organisms), implies that very often scientists cannot provide any useful input to the social debate
without interacting with the rest of the society, as well as the rest of the society cannot perform any
sound decision making without interacting with the scientists [21,26]. That is, the question of ‘how to
improve the quality of a policy process’ must be put, quite quickly, on the agenda of scientists,
decision makers and indeed of the whole society. This extension of the ‘peer community’ is essential
for maintaining the quality of the process of decision making when dealing with reflexive complex
systems together with uncertainty and value conflicts. The new epistemological framework called
‘Post-Normal Science’ was proposed exactly to better deal with two crucial aspects of science in the
policy domain: uncertainty and value conflict.
Post-Normal Science can be characterized in relation to other, complementary scientific strategies,
according to the diagram given in Fig. 5, which is based on two axes: systems uncertainties (on the
descriptive side) and decision stakes (on the normative side).
When both uncertainty and stakes are small, we are in the realm of ‘normal’ academic science,
where it is safe to rely on ‘codified expertise’ without much discussion. When the task is to design and
build a standard elevator, any good practitioner can do it safely, as long as, the codified know—how is
applied properly.
Fig. 5. The PNS graph in relation to semiotic closure—after giampietro [21].
M. Giampietro et al. / Energy xx (2005) 1–28 25
DTD 5 ARTICLE IN PRESS
When either uncertainty or stakes are in the medium range, then the application of routine
techniques and standardized and generalized knowledge is no longer enough. In these cases, skill,
judgment, sometimes even courage are required to adjust the ‘general knowledge’ available to ‘special
situations’. Funtowicz and Ravetz call this ‘professional consultancy’, with the examples of the
surgeon or the senior engineer facing a critical situation. In this situation, the client must have a say in
the choice of the surgeon or the senior engineer that with their choices will determine the final
outcome.
Finally we arrive to cases, in which the possible outcomes are not completely determined by
scientific facts; inferences will (naturally and legitimately) be conditioned by the values held by the
agent. When stakes are very high (as when an institution is seriously threatened by a policy) then
partisan discussion and a defensive tactic will involve challenging every step of a scientific argument
taking sides. An example of this strategy could be the firm denial of the existence of a problem of
global warming by those actors that do not want to implement precautionary policies. We are now in
the realm of Post-Normal Science.
In this situation, the tactic of fighting over the definition of relevant facts or the reliability of
proposed data (e.g. again the study of Stirling and the data of Table 1) should not be considered wrong.
On the contrary, within the realm of Post-Normal Science legitimate contrasting views, also when
these views are held by scientists, have to be openly used to challenge scientific arguments. Taking
side is wrong only when is conducted covertly, as by scientists who present themselves as impartial
judges when, in reality, they are actually committed advocates of a particular view.
4.6. Conclusions
According to the paradigm of Post-Normal Science scientific inputs developed in relation to the
topic of science for governance should no longer be based on mono-criteria analyses generated by
closed committees of experts. Rather multi-criteria analysis of sustainability should be obtained
through participatory procedures of integrated assessment. This requires the adoption of innovative
analytical approaches.
We do not claim that the tools mentioned in this paper for checking the quality of the information
used on the normative and the descriptive side are the best tools available for this task. The tools
mentioned here are those that were presented in the session of Porto Venere to which this paper refers
to. In any case, we believe that they make the point that it is possible to adopt innovative approaches to
do things in a different way.
If we accept the idea that the process of decision making is intended as a search for ‘satisfying
solutions’, the new role of scientists should be that of facilitating the negotiation among stakeholders
by clarifying the nature and possible consequences of trade-offs in relation to non-equivalent criteria
of quality and in face of uncertainty on predictions. The alternative is hampering such a process by
taking sides on substantive basis. That is, by indicating the ‘best’ solution among the considered
options following a process that in any case will be reflecting either vested interests or personal views.
The implications of complexity in relation to science for governance entail a two-way dialogue of
science within the society: scientists have to teach and learn at the same time. This is in contrast with
the concept of substantive rationality that entails a one-way flow of information. The concept of Post-
Normal Science puts back the scientists within the continuous process of social learning, rather than
holding science as something external and above it.
M. Giampietro et al. / Energy xx (2005) 1–2826
DTD 5 ARTICLE IN PRESS
References
[1] Brown MT. Final document of the International Workshop held in Porto Venere. In: Ulgiati S, Brown MT,
Giampietro M, Herendeen R, Mayumi K, editors. Advances in energy studies. Exploring supplies, constraints, and
strategies. Padova, Italy: Modesti Publisher; 2001. p. 305–18.
[2] Allen TFH, Tainter J, Pires C, Hoekstra TW. Dragnet ecology—“just the fact ma’am“: the privilege of science in a
postmodern world. BioScience 2001;51:475–85.
[3] Anderson JL. Embracing uncertainty: the interface of Bayesian statistics and cognitive psychology. Conservation
Ecology [online] 1988;2(1):2. Available on internet URL: http://www.consecol.org/vol2/iss1/art2
[4] Meffe GK, Viederman S. Combining science and policy conservation in biology. Wildlife Soc Bull 1995;23(3):327–32.
[5] Walters CJ. Adaptive management of renewable resources. New York: MacMillan; 1986.
[6] Clark TW. Creating and using knowledge for species and ecosystem conservation: science, organizations, and policy.
Perspect Biol Med 1993;36(3):497–525.
[7] Brunner RD, Clark TW. A practice-based approach to ecosystem management. Conservation Biol 1997;11(1):48–56.
[8] Weeks P, Packard JM. Acceptance of scientific management by natural resource-dependent communities. Conservation
Biol 1997;11(1):236–45.
[9] Funtowicz SO, Ravetz JR. A new scientific methodology for global environmental issues. In: Costanza R, editor.
Ecological economics. New York: Columbia; 1991. p. 137–52.
[10] Pattee HH. Evolving self-reference: matter, symbols, and semantic closure. Commun Cognit Artif Intell 1995;12:9–28.
[11] Kuhn TS. The structure of scientific revolutions. Chicago: University of Chicago Press; 1962.
[12] Ravetz JR, Funtowicz SO. (Guest editors) Special Issue of Futures dedicated to Post-Normal Science [Futures: 1999.
vol. 31].
[13] Ashby WR. An introduction to cybernetics. London: Chapman & Hall Ltd; 1957.
[14] Rosen R. Anticipatory systems: philosophical, mathematical and methodological foundations. New York: Pergamon
Press; 1985.
[15] Rosen R. Essays on life itself. New York: Columbia University Press; 2000.
[16] Kampis G. Self-modifying systems in biology and cognitive science: a new framework for dynamics, information and
complexity. Oxford: Pergamon Press; 1991.
[17] Georgescu-Roegen N. The entropy law and the economic process. Cambridge: Harvard University Press; 1971.
[18] Box GEP. Robustness is the strategy of scientific model building. In: Launer RL, Wilkinson GN, editors. Robustness in
statistics. New York: Academic Press; 1979. p. 201–36.
[19] Allen TFH, Starr TB. Hierarchy—perspectives for ecological complexity. Chicago: The University of Chicago Press;
1982.
[20] Ahl V, Allen TFH. Hierarchy theory. New York: Columbia University Press; 1996.
[21] Giampietro M. Multi-scale integrated analysis of agroecosystems. Boca Raton: CRC Press; 2003.
[22] Giampietro M. Complexity and scales: the challenge for integrated assessment. In: Rotmans J, Rothman DS, editors.
Scaling issues in integrated assessment. The Netherlands: Swets & Zeitlinger B.V. Lissen; 2003.
[23] Giampietro M, Ramos-Martin J. Multi-scale integrated analysis of sustainability: a methodological tool to improve the
quality of narratives, Int J Global Environ Issues [in press].
[24] Giampietro M, Mayumi K, guest editors. Societal metabolism and multiple-scales integrated assessment. Popul Environ
2000;22(2):95–254.
[25] Giampietro M, Mayumi K, guest editors. Societal metabolism and multiple-scales integrated assessment. Popul Environ
2001;22(3):255–352.
[26] Munda G. Social multi-criteria evaluation (SMCE): methodological foundations and operational consequences. Eur
J Oper Res 2004;158(3):662–77.
[27] Carnot S. Reflexions sur la puissance motrice du feu sur les machines propres a developper cette puissance. Paris:
Bachelier, Libraire; 1824.
[28] Scott MJ, Antonsson EK. Aggregation functions for engineering design trade-offs. Fuzzy Sets Syst 1998;99(3):253–64.
[29] Checkland P. Systems thinking, systems practice. Chicester: Wiley; 1981.
[30] Checkland P, Scholes J. Soft-systems methodology in action. Chicester: Wiley; 1990.
M. Giampietro et al. / Energy xx (2005) 1–28 27
DTD 5 ARTICLE IN PRESS
[31] Roling N, Wagemakers A, editors. Facilitating sustainable agriculture participatory learning and adaptive management
in times of environmental uncertainty. Cambridge: Cambridge University Press; 1998.
[32] Giampietro M, Mayumi K. Complex systems and energy. In: Cleveland C, editor. Encyclopedia of energy. San Diego:
Elsevier; 2004.
[33] Fluck RC. Net energy sequestered in agricultural labor. Trans Am Soc Agric Eng 1981;24:1449–55.
[34] Fluck RC. Energy of human labor, in Energy in Farm Production, vol. 6 of Energy in World Agriculture, Fluck, RC.
Amsterdam: Elsevier; 1992.
[35] Giampietro M, Pimentel D. Assessment of the energetics of human labor. Agric Ecosyst Environ 1990;32:257–72.
[36] Giampietro M, Pimentel D. Energy efficiency: assessing the interaction between humans and their environment. Ecol
Econ 1991;4:117–44.
[37] Giampietro M, Pimentel D, Cerretelli G. Energy analysis of agricultural ecosystem management: human return and
sustainability. Agric Ecosyst Environ 1992;38:219–44.
[38] Giampietro M, Bukkens SGF, Pimentel D. Labor productivity: a biophysical definition and assessment. Human Ecol
1993;21(3):229–60.
[39] Salthe S. Evolving hierarchical systems: their structure and representation. New York: Columbia University Press;
1985.
[40] Gomiero T, Giampietro M. Overview of graphic tools for data representation in integrated analysis of farming systems.
Int J Global Environ Issues [in press].
[41] Stirling A. Multicriteria mapping: mitigating the problems of environmental evaluation?. In: Foster J, editor. Valuating
nature: economics, ethics and environment. London: Routledge; 1997.
[42] Munda G. Multicriteria evaluation in a fuzzy environment. Theory and applications in ecological economics.
Heidelberg: Physica-Verlag; 1995.
[43] Commission of the European Communities. Communication from the Commission on the Precautionary Principle,
02.02.2000. COM(2000)1. Brussels: The Commission, 2000. See also: http://europa.eu.int/comm/off/com/health_
consumer/precaution.htm
[44] Simon HA. From substantive to precedural rationality. In: Latais JS, editor. Methods and appraisal in economics.
Cambridge: Cambridge University Press; 1976.
[45] Giampietro M. The precautionary principle and ecological hazards of genetically modified organisms. Ambio 2002;
31(6):466–70.
[46] Abbott EA. Flatland, a romance of many dimensions. Boston: Little, Brown & Co.; 1935.
[47] Munda G, Nijkamp P. Policy analysis for sustainable development: an operational approach to natural
park management. In: Coccossis H, Nijkamp P, editors. Planning for our cultural heritage. Aldershot: Avebury;
1995. p. 69–88.
[48] Munda G, Nijkamp P, Rietveld P. Qualitative multicriteria methods for fuzzy evaluation problems: an illustration of
economic-ecological evaluation. Eur J Oper Res 1995;82:79–97.
[49] Munda G. Multicriteria evaluation as a multidimensional approach to welfare measurement. In: van den Bergh J, van
der Straaten J, editors. Economy and ecosystems in change: analytical and historical approaches. Cheltenham, UK:
Edward Elgar; 1997. p. 96–115.
[50] Munda G, Paruccini M, Rossi G. Multicriteria evaluation methods in renewable resource management: the case of
integrated water management under drought conditions. In: Beinat E, Nijkamp P, editors. Multicriteria evaluation in
land-use management: methodologies and case studies. Dordrecht: Kluwer; 1998. p. 79–94.
[51] De Marchi B, Funtowicz S, Lo Cascio S, Munda G. Combining participative and institutional approaches with
multicriteria evaluation. An empirical study for water issues in Troina, Sicily. Ecol Econ 2000;34(2):267–82.
Glossary
Assessment: a critical evaluation and analysis of information relevant for decision making.
Integrated Assessment: the simultaneous appraisal of attributes of performance referring either to different
M. Giampietro et al. / Energy xx (2005) 1–2828
DTD 5 ARTICLE IN PRESS
dimensions of analysis and/or different scales. It requires the simultaneous use of indicators developed in
different disciplinary fields.
Multi-criteria space: an integrated set of logically independent criteria used to characterize the performance of a
system in relation to a selected set of policy options.
Autism: a variable developmental disorder that appears by age three and is characterized by impairment of the
ability to form normal social relationships, by impairment of the ability to communicate with others, and by
stereotyped behavior patterns [Merriam-Webster Dictionary].
Epistemology: the study or a theory of the nature and grounds of knowledge especially with reference to its
limits and validity [Merriam-Webster Dictionary].
Problem structuring: the result of a process of formalization of a perceived problem in scientific terms. It
includes several steps: issue identification, choice of criteria, choice of indicators, choice of models, choice of
appropriate protocols for data collection.
Semiotic closure: a successful action in the reality which is reflecting a successful analysis and planning.
This expression coined by Pattee (1995) is related to the concept of Semiotic Triad introduced by Peirce
[/APPLY/REPRESENT/TRANDSUCE/APPLY/]—Peirce (1935).
Scale: the relation between a given perception and representation of the reality, which is determined by the
choice made by the observer about the relevant attributes associated with the identity of the observed system.
The finite ability to process information of any observer/agent implies that after having defined the identity of
the observed system both the relative perception and the representation will be characterized by a given grain
(minimum detectable gradients, differentials) and a given extend (size of the domain of observation).
Non-equivalent descriptive domains: representations of the reality based on definitions of space and time that
are not reducible to each other. An example of two non-equivalent descriptive domains is given by two
pictures of the same person taken by: (1) a microscope and (2) a regular camera. With a microscope it is
impossible to perceive the quality ‘face’ observable only when using a regular camera.
Technical incommensurability: it is impossible to reduce to a single model and data set heterogeneous
information referring to representations belonging to non-equivalent descriptive domains.
Social incommensurability: in policy problems socials actors always call a set of contrasting and legitimate
values, perceptions and interests. This implies that any decision is always associated with the generation of
winners and losers.
Hierarchy Theory: a theory of the observer’s role in any formal study of complex systems (Ahl and Allen, 1996,
p. 29).
Triadic Filtering: the operation required to define an identity for the observed system. It implies determining an
expected relation among types used to represent on three contiguous hierarchical levels: (1) the whole; (2) its
parts; (3) the associative context (admissible environment).
Stakeholders: those affecting and affected by the policy (or changes) under analysis.
Jevon’s paradox: an increase in efficiency in a using a resource leads, in the medium to long term, to an
increased consumption of that resource (rather than a decrease). For complex adaptive systems ‘ceteris’ are
never ‘paribus’.