How We Ought to Think of “Ought” page 1

How We Ought to Think of “Ought”

I Introduction

The present essay offers a unified account of the modals

“ought,” “must,” and “is obliged to.” It argues that there are no

special moral senses of “ought” and “must.” The same predicate

may be normative in application to agents, but non-normative in

application to non-agents.1 The central theses are: 1) that

“ought” and “must” are each single, univocal modal predicates

with very wide application; 2) that “ought” is a different modal

predicate with different truth-conditions from the truth-

conditions of “must” and “is obligated to.”

This account is relevant to debates in at least two areas of

philosophy: In ethics, because much of the time what we ought to

do we are also obligated to do, and most of the time, for some

kinds of obligations, we ought to do what we are obligated to do,

“must” and “ought” are easily confused. In virtue epistemology,

there is a long-standing debate about how it can be that we ought

to believe things.2

1 Matthew Chrisman (2012), On the Meaning of “Ought” Studies in

Metaethics, vol.7

has excellent arguments for such a unification of the various

“senses” of “ought.” 2 See Chrisman, Matthew. (2008). Ought to Do. THIS JOURNAL vol

CV, 346-370.

How We Ought to Think of “Ought” page 2

Section II presents some semantic and logical facts about

“ought,” “must” and “is obligated to.” Sections III offers

theories of the truth-conditions of these modals which fit the

data in Section II. Section IV very briefly sketches some

consequences of this account for ethical theory.

My apparatus differs from that employed by most

theoreticians of modalities. Modals are not quantifications over

possible worlds but rather relations between propositions and

sets of propositions. The semantics of modals says what those

predicates are and how many places they have. The theory of

modals is an account of what has to obtain for a particular modal

predicate to be true of a proposition relative to a set of

propositions.

The treatment of modals as predicates simplifies in at least

four ways:

First, it assimilates the treatment of adverbial

modification of modals to ordinary adverbs. “Physically” in

“physically necessary” is the same predicate as it is in

“physically ugly.”

Second, it makes it relatively easy to give accounts of

complex modal expressions. “Ought very much to” fits with one’s

account of comparative adjectives if “ought” is a predicate.

“Slightly more probable than” likewise is not a special case if

“probable” is an adjective.

Third, some modals are clearly relations of a proposition to

a set of propositions. “One must register motor vehicles”

How We Ought to Think of “Ought” page 3

obviously relates “One registers motor vehicles” to a body of

laws. An account of all modals as relations between propositions

and sets of propositions thus treats modals uniformly.

For the modals that this essay discusses, “ought” “must”

“is obligated to” and “reasonably expected that,” it is simpler

just to describe the relationship a given modal predicate

expresses rather than to construct the appropriate condition on a

set of possible worlds. While heuristic models in terms of

possible worlds may illuminate some aspects of our understanding

of some modals, such heuristics do not seem to be relevant to our

understanding of all modals. For modals such as “somewhat more

likely than” and “only barely possible” the heuristic value of

thinking of the modals as quantifications over possible worlds is

obscure.

“Necessary” and “possibly” are special limit cases analogous

to the quantifiers “all” and “some.” Because “all” and “some” are

extreme quantifiers, unlike “ few” and “most,” it is possible to

represent “All frogs are green” and “Some frogs are green” as

operators on truth-functions. But provably “Most frogs are green”

cannot be so represented.3 Just as logicians were misled by

special features of the limit quantifiers “all” and “some” to

3 See Jon Barwise and Robin Cooper (1981).Generalized Quantifiers

and Natural Language. Linguistics and Philosophy 4. 159-219, citing

Nicholas Rescher (1964). Plurality Quantification. The Journal of

Symbolic Logic, vol. 27. 373-74..

How We Ought to Think of “Ought” page 4

take them to be operators,4 so contemporary linguistic theorists

are misled by the success of possible world explications of the

extreme limit modalities “is necessary” and “is possible” to

suppose that modals are quantifications over possible worlds.

Just as quantifiers are really predicates of sets, so modals are

really relations between propositions and sets of propositions.

II Some features of “ought” “must” and “obligated”

a) Non-agent subjects of “ought” and “must”

Some “ought”-sentences have nothing to do with actions by

agents. For instance, the asparagus ought to be up by now, since

it is mid-May. Asparagus has no obligations. So, at least in

these non-agent applications, “ought” means something different

from “is obligated to.” Some of these non-agent applications of

“ought” make perfect sense, but a different sense, with “must”

substituted for “ought.”

These non-agent applications of “ought” and related words5

are unlikely to be metaphorical or to be homonyms of modals that

only apply to agents. The hypothesis that we have families of

homonyms should be a last resort, rather than what we immediately

accept as the right theory.6

4 Heim, Irene and Kratzer, Angelika (1998). Semantics in Generative

Grammar. (Oxford: Blackwell Publishing) page 191.

5 There is more to say about “should,” which, besides being

virtually synonymous with “ought to,” is related to “would” and

“could,” with their relatives “will” and “can.”

How We Ought to Think of “Ought” page 5

In the next section, I propose a way to treat “ought” as

univocal throughout its applications to agents and to non-agents

and propose a similar account of “must.” “What goes up must come

down,” “You must move your king when it is in check,” and “You

must not lie” are the same “must.”

b) Obligations, requirements, and “ought”

“If you major in Philosophy, you ought to take Philosophy

2211” implies that the course is recommended but not required. On

the other hand, “If you major in philosophy you must take

Philosophy 2211” describes a requirement. My accountant’s advice,

“If you make over $100,000 per year you ought to invest part of

your income in an IRA” is quite different from “If you make over

$100,000 per year you are obligated to fill out the long form.”

Conditional “ought”-sentences do not warrant some of the

inferences warranted by conditional obligation sentences. The

guidelines may say both “A student who majors in philosophy ought

to take Philosophy 2211” and “A student who has taken Math 1101

ought not to take Philosophy 2211.” But a student who majors in

philosophy and has taken Math 1101 is not someone who both ought

and ought not to not take Philosophy 2211.

Conditional obligation sentences are obligations that a

conditional be true and true antecedents yield categorical

obligations. Thus the “must” analogs of the above “ought”-

sentences, “A student who majors in philosophy must take

6 Matthew Chrisman (2012), On the Meaning of “Ought”. Studies in

Metaethics, vol.7, makes this point.

How We Ought to Think of “Ought” page 6

Philosophy 2211,” and “A student who has taken Math 1101 must

not take Philosophy 2211,” forbid someone who has taken Math 1101

from majoring in philosophy.

As Davidson7 observes, “ought”-sentences behave in much the

way that conditional probability sentences do. Briefly,

conditional probabilities with true antecedents do not yield

categorical probabilities. The truths “Anyone against gay

marriage is probably a Republican” and “Anyone who voted for

Obama is probably not a Republican” do not jointly warrant the

conclusion that Fred, who is against gay marriage and who voted

for Obama, is probably both a Republican and not a Republican.8

Many modals have the same non-detachability feature. In

particular, from “if A it is reasonable to believe that B,” and

“A” it does not follow that it is reasonable to believe that B.

To see this, replace “probably” with “reasonable to believe that”

iin the example above. “If A it is not surprising that B” has the

same feature. I argue below that the reason these modal

conditionals do not detach is that they are relations between

propositions and sets of propositions. The “if”-clause adds an

element to the set. The consequent is then second argument in a

relation to that enlarged set.

7 Davidson, Donald. (1970). How is Weakness of the Will Possible.In Davidson (1980). Essays on Actions and Events. (Oxford UniversityPress) 21-42.

8 Hempel, Carl. (1960). Inductive Inconsistencies. Synthese Volume 12 Number 4. 439-469 makes this point..

How We Ought to Think of “Ought” page 7

The “ought” in a conditional “ought”-sentence likewise

characterizes some kind of bearing the antecedent has in relation

to the consequent,9 relative to some considerations. If “ought”-

sentences also have form analogous to modals like conditional

probability, the “antecedents” of conditional “ought” sentences

describe the bearing of the consideration in the antecedent on

the consequent. I discuss that bearing below.

c) Anankastic “ought” and “obligation” sentences

Conditional “ought”-sentences with an agent’s desire

described in the antecedent make recommendations, whereas such

conditional “must”-sentences state requirements. “If you want a

hamburger, you ought to try Joe’s Grill” makes a recommendation

which is not a requirement. On the other hand, “If you want a Big

Mac, you must go to MacDonald’s” describes what is required in

order to satisfy your desire.

9 In modals other than the extreme ones, “antecedent” and

“consequent” are misleading. David Lewis proved that conditional

probability cannot be the categorical probability of any

conditional. (Lewis, David. (1976) Probabilities of Conditionals

and Conditional Probabilities. Philosophical Review 85 (3). 297-

315.) “If” marks the conjectural addition to the modality base.

How We Ought to Think of “Ought” page 8

As several have noted,10 the anankastic conditionals with

“ought” have multiple understandings. Both “If you want heroin,

you ought to contact Fred” and “If you want heroin, you ought to

enter a drug rehabilitation program” can be understood as truths.

The possibility of multiple readings suggests that “ought”-

sentences are relative to sets of considerations.

c) Peculiarities of “must”

“Must,” although often inter-changeable with “is obliged to”

applies much more widely. It applies to non-agents, as in “What

goes up must come down.” “Must,” although akin to “is necessary”

is actually sometimes weaker than “is.” “Must” implies an

inference from something. When I say “This must be Spring

Street,” relying on the guy at the gas station, “This is Spring

Street” would convey more confidence.

d) Prima facie obligations

Many ethicists explain some of the differences between

“ought” and “obligation” by the concept of “prima facie

obligation,”11 that many “ought”-principles are ceteris paribus

principles, but still principle of obligation. This idea is

10 For instance, Von Fintel, Kai and Iatridou, Sabine,(2008). How

to Say Ought in Foreign: The Composition of Weak Necessity

Modals, in Gueron and Lecarme, (ed.s) Time and Modality. (Springer),

115-41.11 Ross, W. D. (1930). The Right and the Good. (Oxford University

Press.)

How We Ought to Think of “Ought” page 9

inadequate. First, it applies only to “ought”s which involve

agency. Defeasible weather-principles need some other account.

Second, the resemblance of “ought”-sentences to conditional

probability sentences argued for by Davidson12 suggests that the

relationship between the antecedents of conditional “ought”

sentences and their consequents is unlike that between

antecedents and consequents of “must” and “is obligated to”

sentences. Third, there is no plausible way to treat normal

anankastic “ought”-sentences as obligations of any sort.

The next section replaces “prima facie obligation” with an

account that respects the linguistic and logical differences

between “ought” and “is obligated.” .

III Theories of “ought,” “must,” and “is obligated”

a) Preliminary: Consideration-bases

A consideration-base13 for a modality is a set of

propositions. “Ought”, “must,” and “is obligated to” are two-

place relations between such bases and the proposition in the

content clause. What consideration-base a speaker means is guided

12 Davidson, Donald. (1970). How Is Weakness of the Will

Possible?. In Davidson (1980). Essays on Actions and Events. (Oxford

University Press) 21-42.

13 I use “consideration-base” rather than “modal base” because

the concept is somewhat different from that used by Angelika

Kratzer (2012) Modals and Conditionals. (Oxford University Press)

and others. Consideration-bases are sets of propositions.

How We Ought to Think of “Ought” page 10

by, but not determined by, context. “Ought” and “must” both have

consideration-bases, but these bases have different requirements.

These modals also express different relations between the base

and the content clause.

Consideration-bases can be referred to without being listed

and without the speaker having a criterion for whether a given

proposition is in the set. So, “Fred’s self-interest,”

“Connecticut statutes,” “how things are,” and “the laws of

nature” all refer to sets of propositions. This is obvious with

consideration-bases associated for “must” that are

uncontroversially sets of propositions. For “You must not turn

left at red lights” the usual consideration-base is traffic laws.

But few can quote the relevant statute.

Consideration-bases for true “ought” sentences are sets of

truths--the data relative to which the proposition following

“ought” is evaluated. There are different consideration-bases for

a given “ought”-sentence, and different truth-values relative to

those different consideration-bases. The relation between a

consideration-base and the proposition following “ought” is

“idealized reasonable expectation,” explicated below. Conditional

“ought” sentences add the proposition in the antecedent to the

consideration-base and evaluate the support that that enlarged

base would give to the proposition following the “ought.”

Consideration-bases for “must” and “is obligated” sentences

are also sets of propositions, but they need not be truths.

Distinct consideration-bases can be law codes, rules of chess,

How We Ought to Think of “Ought” page 11

information from tour schedules, and laws of nature, along with

information about particular circumstances. While these

consideration-bases may be expressed as imperatives, their

content, is a proposition with a truth-value. So, “Thou shalt not

kill,” as a component of a consideration-base, is a false

universal description of the behavior of those to whom it

applies. None of them kill. The relation between a given

consideration-base and the “must” sentence is “consequence.”

Conditional “must” sentences thus can be treated indifferently

either as conditionals that are consequences of the

consideration-base, or as adding the antecedent of the

conditional to the consideration-base. That is, when “if A then

you must make it the case that B” is true, then so is “You must

make it the case that if A then B.”

There are several possible consideration-bases for any given

“must” or “ought” sentence. We interpret a modal utterance by

assigning some consideration-base. That assignment is guided, but

not infallibly, by context and plausibility. The speaker’s

intention fixes the consideration-base.

Unusual and unlikely bases may be intended, just as

improbable readings may be correct. One of the syntactically

possible readings of “My dog stays at a clean, expensive pet

kennel” implies that this kennel is for clean, expensive dogs.

Since such kennels are unlikely, we are unlikely to assign that

reading. But a speaker could mean that. An utterance of “You must

not kill people” likewise has several possible consideration-

How We Ought to Think of “Ought” page 12

bases. Struck by Hagar’s tender demeanor, I remark to him “You

must not kill people.” A speaker can intend a base whether or not

context makes it salient. A speaker who wants to communicate may

use an adverb such as “logically,” “morally,” or “for his own

sake” to make the intended consideration-base clear.

The requirement for being possibly understood means that not

just any set of propositions can be a consideration-base. Given

that language use is primarily communication, only a

consideration-base that another can grasp is a possible

consideration-base.

b) “Ought”

If the same modal applies to non-agents and agents,

something about non-agent attributions also applies to agent

attributions. The obvious feature of non-agent “ought” is that,

relative to the consideration-base, the content clause is also to

be reasonably expected, another modal. What the weather ought to

be tomorrow and how far the flood waters ought to have progressed

are equally well described as what can be reasonably expected,

given the consideration-base, the relevant facts.

For this relation to provide the link between non-agent

“ought” and “ought”s that apply to agents, the modals “ought” and

“reasonably expected” have to be connected and there must be

reasonable expectations of an appropriate kind about agents. I

propose that “ought” holds between a consideration-base and a

proposition just in case there is a chain of reasonable

expectations from the considerations in the base to the

How We Ought to Think of “Ought” page 13

proposition. I propose further that interpreting entities as

rational agents is applying reasonable-expectations principles

from consideration-bases that take into account interpretive

principles about agents.

b1) “ought” as a chain of reasonable connections relative to a

consideration-base

For the first part of the link, I explicate “ought” as an

idealization of “reasonably expected relative to a consideration-

base.” By an idealization, I mean a chain of conditional

reasonably expected propositions, relative to a consideration-

base, such that the consequent of the element n is the antecedent

of element n+1. For example, relative to a consideration-base

consisting of truths about the sums and products of one-digit

numbers an agent, as a believer in truths, can reasonably be

expected to get the right answer to any such multiplication. With

that consideration-base, there is a chain of reasonable

expectations such that the result of multiplying two seven-digit

numbers ought to be the arithmetically correct result. So,

relative to just the truths of arithmetic, and the reasonable

expectation that an agent will get simple calculations right, the

agent ought to get the arithmetically correct result.

However, we cannot reasonably expect a human calculator to

get the mathematically right result, for a very long calculation.

Rather, relative to a consideration-base that abstracts from

human imperfection, there is a chain of reasonable expectations

to the mathematically right result.

How We Ought to Think of “Ought” page 14

If the consideration-base includes the further information

that the agent is a fourth grader new to multiplication, we can

still reasonably expect him to get the right answer only most of

the time when multiplying or adding one-digit numbers. After many

such calculations, it will be true that he ought to have made a

mistake by now. So, the number he ought to have come up with

ought to be different from the correct mathematical answer. Since

he is both an agent and a fourth-grader, both “ought”-sentences

are true, relative to different consideration-bases.

An idealization of “can be reasonably expected to,” relative

to a consideration-base, is what would result from a sequence of

events if what could reasonably be expected to happen at each

stage in the sequence, relative to the consideration-base,

happened at every stage. If the chain is very short, as with “If

you turn the key, the car ought to start,” “ought” and

“reasonably expected that” can amount to the same thing relative

to a base.

The difference between “ought” and reasonable expectation

relative to a consideration-base is that an idealization can

involve very long chains of steps. Since a reasonable expectation

that p and a reasonable expectation that q if p need not imply a

reasonable expectation of q, chains of reasonable expectations

result in diminished reasonable expectation. Many modals have the

same feature, such as “it is not surprising that” and “probably.”

The result of the diminution is that we cannot reasonably expect

an agent to get the correct mathematical result in a long

How We Ought to Think of “Ought” page 15

calculation, even relative to a consideration-base that the agent

is an agent, and so inferring truths, and the truths of

arithmetic. What ought to be may differ from what is reasonably

to be expected.

Since “ought” draws consequences for the consequent relative

to a consideration-base, a base can take special circumstances

into account. Thus, lottery paradox problems do not arise. Our

reasonable expectation, of each lottery ticket, that it will not

win will not lead to the conclusion that it ought to be that no

ticket wins, because the proposition that someone wins is part of

an appropriate consideration-base. The chain would in effect

build up a conjunction of claims that individuals will not win.

The fact that it is a lottery blocks the “ought”-claim that

everyone in the conjunction fails to win. One of these tickets

ought to have won. That is, each of the reasonable expectations

is a reasonable expectation relative to the consideration-base.

In this special circumstance, the consideration-base includes the

fact that these are tickets in a lottery. So, even though of each

individual, we can say both that it is reasonable to expect it to

lose and that it ought to lose, we do not get the result that

they ought to all lose.

This is parallel to the chain of reasonable expectations in

the case of the fourth grader. Just as it begins to be true that

he ought to have made a mistake by now, so it begins to be true

that one of the people we have considered ought to have won. The

lottery imposes an addition to the consideration-base that

How We Ought to Think of “Ought” page 16

requires that the fact that it is a lottery be part of the

consideration-base at every step, and eventually becomes relevant

to whether an “ought”-sentence is true.

b2) reasonable expectations of agents as rationality

Just in virtue of an entity’s being an agent, we can

reasonably expect that entity to be aware of objects in its

environment, to reason as we do, and want things we want. These

elementary reasonable expectations of agents are, for instance,

part of Davidson’s account of interpretation. In Davidson’s 14

phrase, we interpret agents’ behavior by making them maximally

“believers of the true and seekers of the good.”

Maximization principles of interpretation are thus

reasonable expectations with the consideration base including “is

an agent.” Some applications of these principles are more

reasonably expected than others. Basic beliefs, such as that

there are physical objects, and basic desires, such as dislike of

intense pain, are ascribed to agents with great warrant. Other

desires and beliefs, such as a desire to play contract bridge or

the belief that Augusta is the captal of Maine, are not so

central. A consideration-base including differences in perceptual

history explains why some of our beliefs are not shared by

another agent. But, relative just to the consideration-base that

14 On page 222 of Donald Davidson, (1970). Mental Events. In

Davidson (1980). Essays on Actions and Events. Oxford University Press.

207-227.

How We Ought to Think of “Ought” page 17

the other is an agent, what we believe and want is reasonable to

assign to the other agent. Thus, even though we do not expect

another agent to know what we know, there is a consideration-base

restricted to just the agency of the other, what we take to be

the truths, and our beliefs and desires, from which a chain of

reasonable expectations follows. “She ought to flee, because

the flood waters are coming,” can be true even though she has no

way of knowing about the flood. That consideration-base rests on

the reasonable expectation, relative just to being an agent, she

believes what we believe about the flood waters. Relative to

another base including her epistemic position relative to flood

information, she ought to stay at her desk.

What Fred ought to do may be relative to a consideration-

base which includes preferences peculiar to Fred. The “ought”-

sentence with that base may differ from what Fred morally or

prudentially ought to do. “Fred ought to be very drunk by now.”

On the other hand, the consideration-base which leaves out Fred’s

self-interest, odd preferences, and weaknesses and just includes

what Fred can be expected to want just given that he is a

rational agent, provides starting points for a chain whose last

element is a candidate for what Fred morally ought to do.

The notion of rationality constrains contents of desires as

well as the structure of practical inference. Interpretation of

action is only possible given such content constraints. If all

desires were equally rational, then any behavior would have an

optimal interpretation as an intentional action that reflected

How We Ought to Think of “Ought” page 18

perfectly correct beliefs. For instance, if a person drops a

large rock on her foot, the interpretation that she wanted pain

and knew that rocks landing on feet produce pain would be as

acceptable as supposing the event to be accidental. Irrational,

otherwise not-reasonably ascribed desires can be attributed to a

rational agent, but only on the basis of having ascribed other

desires we share and so can understand.

b3) “ought” and consideration-bases

Even in the simple non-agent attributions of “ought” the

same “ought”-sentence may have different truth-values with

different consideration-bases. From a consideration-base

consisting of general truths about small gasoline engines it is

unreasonable to expect that my mower will start will one pull of

the starting cord. However, given that it is a Honda, that is,

adding “This is a Honda” to the consideration-base, it should

start with one pull.

Conditional probability was proven by Lewis15 not to be a

probability applying to a conditional. I am supposing that

“reasonably expected” is a modal like conditional probability in

this respect. Thus, an idealization of “if p, it can be

reasonably expected that q,” that is, “if p, then ought q,” is

not an “ought” applied to a conditional. The bearing of the

15 Lewis, David. (1976). Probabilities of Conditionals and

Conditional Probabilities. Philosophical Review 85 (3). 297-315.

How We Ought to Think of “Ought” page 19

consideration-base on an “ought”-sentence is an idealization of

reasonable expectation.

The “antecedent” of a conditional “ought”-sentence

supplements the consideration-base with another proposition, and

the “ought” is evaluated relative to that expanded set. So, “If

George wants a hamburger, he should go to Joe’s Grill” appraises

the idealized support “George goes to Joe’s Grill” gets relative

to a consideration-base which just includes George’s situation

with respect to hamburger vendors and interpretive maxims about

the connection between desires and action, supplemented by the

supposition that George wants a hamburger. Relative to another

consideration-base which includes George’s long-term self-

interest, his heart condition and the effects of saturated fat,

“If George wants a hamburger, he ought to have something

healthier” might be true.

The links of a chain of reasonable expectations making a

prudential “ought” true are reasonable expectations about what an

agent can reasonably be expected to believe and want. The

prudential “ought” is thus an idealization of someone with good

information and reasonable desires. An anankastic “ought” has an

antecedent adding a particular desire, and the consequent is an

action optimal for a person with reasonable desires, correct

beliefs, and the desire mentioned in the antecedent. Different

bases yield different true anankastic “ought”’s. Someone wanting

heroin ought to see Fred the dealer or get into rehab, depending

on the base.

How We Ought to Think of “Ought” page 20

“Ought”-sentences about an agent’s action thus can be

several ways ambiguous, depending on the intended consideration-

base. Suppose Sue is playing chess with Nero, it’s her move, and

there is a mate in two starting with a bishop sacrifice on f7:

First, if the consideration-base is just chess goals, then

Sue ought to go bishop to f7 check.

Second, relative to a consideration-base that includes Sue’s

welfare, bishop to f7 check is not what she ought to do. Nero may

take losing very badly.

Third, Nero would be a better emperor if he learned

humility, and losing quickly to Sue may bring this about. The

improvement of life in the Empire that would result from having a

humbler emperor, for Sue and everyone else, makes bishop to f7

check worth the risk. This last consideration-base abstracts from

Sue’s partiality to herself and qualifies as a moral “ought.”

Many other bases may apply. “Sue ought to be under

considerable stress at this point” one observer says truly to

another, intending a consideration-base including information

about how people in such dangerous situations feel.

To summarize: There are different kinds of bases for

reasonable expectations, and so for the “ought”s that are their

idealizations. In the above case, the aims of chess, the over-all

aims of self-preserving humans, the interests of humans

generally, and special expectations about particular individuals

yield various truth-values for the same “ought”-sentence. Given

How We Ought to Think of “Ought” page 21

the position on the board, whether Sue ought to go bishop f7

check depends on what consideration-base is intended.

One could object that this account of “ought” is not

“oughty.” By this account, “ought”s derive from our reasonable

expectations about what a rational agent can be expected to do.

Why should this motivate the agent herself?

First, only “ought”s based on consideration-bases including

what it is to be a rational agent motivate. “Ought”s based on

other than rational-agent consideration-bases do not motivate.

Fred has no motivation to be uncomfortable during the interview

just because he ought to be, given his shyness.

Second, “ought”s resting on consideration-bases that include

expectations about intentional behavior and intentional states

should motivate. Since an agent has to conceive of herself as an

agent, the fact that any agent in these circumstances can

reasonably be expected to act in a certain way is relevant to an

agent’s decision-making. If you expect anyone in your

circumstances to do a given action, you do so by interpreting the

person, that is, by maximizing the agreement between you and the

other. So, you ascribe to the other your beliefs, inference-

patterns, and desires, other things being equal. So, to be aware

that taking an action would be reasonably expected in virtue of

being a rational agent is to be motivated. That’s you. These

expectations reflect norms.

On the Davidsonian perspective, normativity is derived from

rational agency. The content of the concept of rational agency is

How We Ought to Think of “Ought” page 22

embedded in our practices of interpreting each other’s actions,

whether speech actions or not. The maxims of interpretation are

reasonable expectations relative to considerations. If the entity

being interpreted is an agent, this reasonable expectation is

normative and is part of ascribing rationality. The Davidsonian

idea thus follows Kant—the normative is grounded in the very idea

of acting for a reason. A difference is that for a Davidsonian,

the very idea of agency is about reasonable expectations, rather

than universal principles, since the observance of a given maxim

is reasonably maximized rather than absolutely required.

c) “Must”

“Must” is a consequence relation between a proposition and a

consideration-base. If the propositions in the consideration-base

are true, “must” amounts to necessitation. If the propositions in

the consideration-base are not true, the proposition may or may

not be true.

The consideration-base for “What goes up must come down” are

“the laws of nature,” and are true. The consideration-base need

only be referred to, not enumerated. When we say “You must not

shoplift” we appeal to a modal base, the laws of our region, but

very few of us can state even one of those laws. We can say what

must be given the laws of nature without having a even a

criterion for what those laws are.

The consideration-bases of which the content-clause of a

“must” sentence is a deductive consequence have a role similar to

that of “ought”-sentences. Rules of chess, laws of nature, and

How We Ought to Think of “Ought” page 23

the like are typical consideration-bases for “must.” “Must”’s

consideration-base and its relation to that base differs from

“ought”’s base and relation to its base in two ways: First,

rather than being a matter of idealized reasonable expectation,

“must”’s content clauses are deductive consequences of the

consideration-base. Second, the elements of the consideration-

bases for true “must” sentences need not be true, while those of

true “ought”-sentences must be true. “All persons do not

understate their income” does not describe the behavior of those

who must not understate their income.

“Must”-sentences are relative to consideration-bases in much

the same way that “ought” sentences are. There are “must”s true

of Sue’s chess game with Nero. Sue must move her pieces at less

than the speed of light. Since this “must” is supported by truths

about the world, Sue will so move his pieces. Sue must move her

bishop only along a diagonal, since that is a consequence of the

rules of chess. Players usually do so move their pieces, but Sue

might decide to move her bishop as if it were a knight.

Since they are only correct if the content clause is

correct, “epistemic” “must”s have true consideration-bases,

namely the relevant facts. When the tourist says, “Since it’s

Tuesday, this must be Belgium” the “must” may be justified by the

tour schedule. The tour schedule, though, is optimistic, and the

bus is still in Holland. So, the tourist is wrong. The “must” in

these sentences says what must be, given the relevant facts. The

tour guide is the warrant from which the speaker justifiably

How We Ought to Think of “Ought” page 24

conjectures that the relevant facts, what is the case, have the

content clause as a consequence.

The “epistemic” aspect of epistemic “must”s is not a matter

of the consideration-base being propositions known. Rather,

epistemic “must”s are asserted on the basis of the speaker’s

inference from known propositions in the consideration-base to

the rest of the consideration-base. So they are weaker than the

non-modal assertion.

Since contents of “must” sentences are consequences of

consideration-bases, conditional “must”-sentences are

conditionals that are consequences of sets of propositions. A

conditional “must” sentence is indifferently a relation of a

consideration-base to a conditional or a relation of a

consideration-base, supplemented by the antecedent, to the

consequent. Therefore given that “If a taxpayer earns more than

$100,000, a long form must be submitted” and “Fred earns more

than $100,000” are both true, “Fred must submit a long form” is

also true. Conditional “must”-sentences support detachment of the

consequent given the truth of the antecedent, unlike conditional

“ought”-sentences.

IV) Moral “ought” and moral obligation

A detailed treatment of what this account of “ought,”

“must,” and “is obligated” would imply about ethics and meta-

ethics is beyond the scope of this essay. This brief last section

suggests a few consequences of this account of these modals.

a) Moral “ought”

How We Ought to Think of “Ought” page 25

Most importantly, if “ought” is the central ethical notion,

ethical reasoning is more akin to induction rather than to

deduction. This means that ethical reasoning to a conclusion C

can, with further information, become reasoning to not-C. Unlike

deduction, inductive arguments can be weakened by additional

premises.

Moral “ought”s are objective if there is a consideration-

base that yields their content. Showing that there is such a base

and discerning what its elements are is the traditional project

of ethicists. A candidate consideration-base would be those

desires that agents can be expected to have just relative to

being agents, together with true beliefs about a situation. That

candidate for the moral “ought” abstracts from partiality. Given

reasonable desires and beliefs, and the impartiality that

abstraction from self-interest requires, a person ought to desire

what is morally correct.

c2) Moral “ought”s without agents

“There ought to be world peace” and other such apparently

moral “ought”s are not about particular agents. A consideration-

base whose elements have consequences about reasonable

expectations whose idealization would yield the content of such

an “ought”-sentence would include facts about agents

collectively.

“Ought to be” sentences are only true of what agents can

bring about. The existence of good fairies is not in the power of

any human agent or group of agents to bring about. Therefore,

How We Ought to Think of “Ought” page 26

“There ought to be good fairies” is false even though that would

be a good thing. We have reasonable expectations that people

generally want what is desirable. Given that a person is in a

position to act on a desire, it is reasonable to expect that the

person will do so. On Davidsonian interpretive grounds, when

something would be a good thing, people in a position to do it

ought to do it. So, for things under the power of people

collectively, or of some subset of people collectively, some of

those things ought to be the case. The moral “ought to be” is

just the moral “ought” without specificity about agency. What

ought to be is what some unspecified agents ought to do or have

done.16

c3) Moral obligation

“Is obliged to” is the special case of the modal relation

“must” which is strictly true only of agents. Obligations are

deductive consequences of consideration-bases which include

truths about circumstances. I have an obligation to report all of

my income. If I am a slave, I have an obligation to obey my

master. If I am playing chess, I have an obligation to move my

king if it is in check.

Among the consideration-bases of “obligation” sentences are

those that deal with human relations. Bases containing

16 The present theory thus agrees with Matthew Chrisman (2011)

“Ought” and Control. Australasian Journal of Philosophy. 1-19 and

Stephen Finlay and Justin Snedegar, Justin. (forthcoming) One

Ought Too Many. Philosophy and Phenomenological Research.

How We Ought to Think of “Ought” page 27

propositions such as “People do not lie” and “People keep

promises” yield obligations for which there are good reasons to

think that, by and large, they coincide with what a person ought

to do. Those reasons are reasons that appeal to rationality—why

it is that a person would, behind the veil of ignorance, choose

such restrictions, or why it is that a person cannot thoroughly

conceive of herself as a rational actor while ignoring the ends

of other agents.

Obligations deriving from such consideration-bases would

coincide with what a person ought to do, relative just to being

an agent, to the extent that there was a rational ground for

expecting that, by and large, people should not lie, should keep

promises, and so forth. The justification of the approximate

truth of such principles would be the demonstration that these

principles were rational. If that could be done, then the

principles of moral obligation would be part of the rationality

that we attribute reasonably to rational agents.

There may be chains of consequences from consideration-bases

for “must” that always conform to what a rational agent ought to

do, all things considered. For instance, Kant’s17 argument that

it is irrational to favor one’s own interests and not take into

consideration those of other agents could be something that any

rational agent ought to accept, given that the agent is a

17 Kant, Immanuel. (1785). Groundwork of the Metaphysics of Morals. In

Kant (1983) Ethical Philosophy, James Ellington, translator. Hackett.

How We Ought to Think of “Ought” page 28

rational agent, as interpretation supposes. Likewise, Nagel’s18

argument that a person motivated by her own future interests is

irrational not to take other agents’ interests as relevant to her

decisions might be sound. In that case, moral principles, in the

strict sense, would be consequences of rationality. Then a few

moral obligations would in fact be universal moral “ought”s—the

guidelines governing the interpretation of rational agents would

turn out to have these consequences. Moral principles, that

directly determine particular actions, though, such as “Do not

lie” and “Keep your promises” would be rules of thumb. Their

application would be mostly true in actual cases, and mostly true

for good reasons having to do with human interaction.

18 Nagel, Thomas. (1970). The Possibility of Altruism. Clarendon Press: Oxford.