Post on 21-Apr-2023
Carnegie Mellon UniversityResearch Showcase @ CMU
Tepper School of Business
1-1983
The Human Side of Robotics: How Worker's Reactto a RobotLinda ArgoteCarnegie Mellon University, argote@cmu.edu
Paul S. GoodmanCarnegie Mellon University
David SchkadeCarnegie Mellon University
Follow this and additional works at: http://repository.cmu.edu/tepper
Part of the Economic Policy Commons, and the Industrial Organization Commons
This Article is brought to you for free and open access by Research Showcase @ CMU. It has been accepted for inclusion in Tepper School of Businessby an authorized administrator of Research Showcase @ CMU. For more information, please contact research-showcase@andrew.cmu.edu.
Published InSloan Management Review.
To appear in the Spring 183 issue of S loan Management Revie'tv
The Human Side of Robotics: How Wor'ker's React to a Robot
Linda Ar'gote Paul. $ .. Goodman David Schkade
Carnegie-Mellon Univer'sity
September 1982 January 198,
?ar'~ial supper:' !'or this s~udy ',.;as provided by ':,he Gr'aduate 3c!1oo1.. ;Jf ::1dt.,;str'ial A.dnin:'str'ation and the Robotics Instit.ute of Car':1egie-·~e:!.:2.on
:]niJe~'si':,y" 'The :la'S r:ecessar :.1:;
-:oncl:.Jsions r'eached in t.he paper' are l"'ef'lec:t :he '1iews of :'he funding
the sllthor'sl and do institutions" ::"1e
::::~:r:cr's '.."ish to th2nk :nem~er's of the o!"'ganizat:'on t.Je stUGted fef' t~ei:
::8o~er'a'c,icn and 3ssi3tance. ;.j'e also ' .... ish ':.0 ':.hank ?ober't S" .. \t.ki:1, i;e.nie:!. =erg, Jenr:is E;:;ple, :'~ar'k richMan, Lar's p" 2ansen, :.(atheri:1e Schi:Jpe!'", Lee S. Spr''Jull. ?obe?""': L Sut'tcn, and ?aul K .. ~'ir':':sht ~-'.'J?", {:.el;:;ful comme!":.ts ,~n
~ar'!..:,=!" dr'a£'"t D1' this paper 'das presented 3'S the ~;eetir:gs of :. ... ,e ,l~.e!"i~?r:
?syc~olJgical ~ssGciasion, ~as~i~gton, JC, ~ug~s~ 19R2,
ABSTRACT
Ihis study examines war'kers' reactions to t.he introduction of robots
in a fae tor-y. The study focuses on understanding workers' psychological
reactions to this new technology and to the manner in which it was
intrcd uced. Horkers repor'ted that both advantages (lower fatigue) and
d isadv an tag es (increased downtime) wer e assoc iated wi t h the int rod uc t ion
of the robot" Over time, worker s' beliefs about robots became mar'e
complex and pessimistic, Production oper'ator's' jobs, as well as their
interaction patterns with other pr'oduction and SUPPOT't war'ker's changed
with the introduction of the robot. Consequences of these changes for
increases in job stress are examined. A set of strategies fer int.roducing
robots in the factory is discussed.
Robots are being used in increasing number's in offices and fact-,ories
throughout the world. However, little is known about row robots affect
either individual workers or the structure. functionin~ and effeotiveness
of organizations., This paper focuses on workers' react.ions t.o the
introduction of a robot in a factory. He examine how workers react to the
robot itself. as well as how various strategies for intr'oducing robots
affect these reactions" This knowledge should enable managers to make
better decisions about the use of r'obots in their or'ganizations,
In general robot.s can be thought of as machines that. can sense, think
and act, in r'epeatab1e cycles. However 1 given the current state of
robotic technology the sensing and lear'ning functions are not well
developed .. Thus, we will view robots as (electromechanical) devices '.-lith
multiple task capability and progr'ammability .. The curr'ent funct.ions of
most robots in U.S. factories ar'e to tr'ansfer material and to do certain
processes such as welding.
Curr'ently. industrial r'obots are in limited use -- estimates range
from 3.5001
to 5.0002
in the U. S .. in 1980. There are, however, man y
reasons to believe that the number' of robots used in this country ·"ill
, 3 lncrease. These reasons include high labor costs. the current emphasis
on pr'oductivit.Yt and technological improvements in capabilities and costs
of ccbots.
Increased interest in the social impact of robotics has accomp;:mied
t~e use of robots, Questions have been asked about: ;;0,\.[ the use of r'obots
:.Jill affect employment 1eve1s4
, which jobs can be perfoT'!:1ed by ;'obot.s J,
3!'":.-:::' what types of educational and tr'8:ining pr'ogr'ams ar'e needed for' worker's
t,.;hose jobs are affected by robots. 6
This study differs from other wor'k on the social impact of r'obotics
by examining tow individual workers react to the introduct.l.on of a r'obot
at. their' factory" Our focus is on understanding wor'ker s' psychological
r'eactions (e .g .. their att.itudes and motivations) to t.he r'Obot and to t.he
manner in which the r'Obot. was int.rod uced" The more positive war'kers'
reactions to robots are. the mor'e likely organizations will experience
posi ti 'Ie economic con se quences such as increa sed prod ucti vity through the
use of robots,
While this study appears to be one of the first that examines the
effect.s of introducing a robot in a factory on individual wor'kers, t.here
is literature on the impact of technological change on individuals and
organi zations .. Several themes emerge fran that literature which should
help our' understanding of the effects of r'obots in the work place" One
theme is the need to take into account the compatibility between an
or'ganization's technology, its structure, and its member's,,? Fail ur'8 to
consider these factors often r'esults in unintended negative consequences
including increased absenteeism, higher accident rates, and decreased
productivity.
Another theme fr'om studies of technological change is that changes in
technology often lead to changes in the job activities of individual
workers. Hhyte8 found that automation increased the extent to which jobs
were mentally demanding. Eli zur 9 and Hann and Hoffman 10 found that
wor'ker s in automated organizations r'epor'ted a greater sense of control and
r-esponsibilit.y than their' counterparts in l.ess 8utcmated organizations"
l"'iof'Ker's in automated organizations often report a greate:' sense of'
h ed· 11
pressure t .. an workers in less automat organizatIons"
work ..
3
fechnological change also affects social interaction patter'ns et
Whyte 12 found that increased automation decr'eased the opportunities
I(JOT' kef's had to inter'act with their' coworkers" ~4illiams and n
Hilliams'
noted that new technologies often cr'eate new demands on suppor't per'sonnel
and require mar'e coor'dination activities bet.i.,reen support and production
per' sonnel •
Another theme from the oT'ganizational change literature is t.hat
worker involvement in the design of change affects worker acceptance of
14 15 and commitment to the change. In their classic study, eoch and French
found that worker par'ticipation in the design of change was 3ssociated
with higher productivity, lower turnover and fewer acts of aggression
against the company. Similarly, Griener 16 and Crockett 17 both stress that
change attempts are more likely to be successful if everyone affected by
the change is involved in its design and implementation.
While this T'eview is not meant to be exhaustive, it indicates t.hat
workers can resist technological change and that the oppor'tunity to
par'ticipate in the change may r'educe levels of r'esistance. 1he review
also indicates that technological change can affect social interaction
patterns on the job. Decreases in the opportunity to interact with others
are generally associated with increases in wer'ker alienation, stress, and
absenteeism. Ihe revie',.; also suggests that ne'tI technologies can change
work activities .. If the change decreases feelings of vaf'iety, autonomy
and challenge, or introduces activities that are incompatible Hith the
h"or'ker S' abilities aDd pr'eferences, it is likely that workers' attitudes
'dill become more negative and their motivational levels will fall.
:,"hile robots may be viewed as another advance in autoMation! we
believe ':.hat tvor'ker s may view robots as qualitatively different from other
fonns of autO'lation, Hor'ker's have been ex po sed to r'obots l,.;i t.h g;lori fi ed
capabilities on television and in the movies., In arldition, a r'obot. often
directly takes the place of a worker. He think these factors combine to
make the introduction of a r'obot a very sal ient and possibly threatening
event. for worker's"
Research Site and Methodology
To understand row wor'kers react. to the introduction of a robot, data
were collected at a manufacturing plant that waS installing its first
robot. The plant had been in operation for approximately ten years. The
primar y technological pr'ocesses in the plant involved the forging and
machining of various metal alloys" The work force at the plant numbered
about 1,000, was nonunion and predominantly blue collar'" The work force
was fairly stable., For example, the average length of service of
employees in the department wher'e the robot was introduced was eight
years" Relationships between labor and management appear ed to be good.
No walkouts or other examples of industrial strife had occurred at the
plant in recent years.
The company had used a fairly comprehensive set of str'ategies t.o
introduce the robot into the plant. lhese strategies included an open
house in which the operation of the robot was demonstrated, talks by the
plant manager, discussions with fir'st-line supervisor's, and notices posted
in t.he cafeter'ia" 1he company had informed employees that a robot was
going to be intr'oduced at the plant about a year befor'e t.he i'''obot "vas
actually put into operation.
Ihe r'ooot :,.jas intr'oduced in a depar'tment in Hhich the '::>?sic
oper'ations wer'e the milling and grinding of bar stock. 7h er ~ '..Jer e
5
approximately ten different oper'ations in the department.
persons ',·lOrked acr'oss the depart.ment's t.hree shifts.
physically arrange in a horseshoe-like configuration,
operated one or more of the milling and grinding machines"
A total of 4r)
Machines were
Each per son
The workflow in the depar tment '.as pr'imar'il y sequent ial. War ker s
moved pr'oducts from one operation to another' by hand. Ther'e was some
flexibility in the order in which products went thr'ough the var iou"
operations. The majority of the pr'oducts went through most of the
operations; however, not ever'y pr'oduct went. through every operation.
There were so:ne buffer inventor'ies between operations.
Ihe r'obot was placed at the beginning of the workflow in the
department, Ihe robot loaded and unloaded two milling machines. One
per'son operated the robot on each shift.
We interviewed production '.orkers on each shift in the depar'tment
where the robot was intr'oduced before and after the robot was put on line"
These pr'oduction workers were our primar'y sample" Inter'views were
conducted with them during two separate visits to the plant in 1981, about
2 1/2 months before and 2 1/2 months after the introduction of the roboL
The individuals we interviewed during the second visit wer'e the same as
those interviewed during the first; however. some workers that
par:,icipated were not available at t.he time of the second due to factor's
such as vacations and illness" During the fir st visit, 37 employees fr'om
the depar'tment in which t.he robot was intr'oduced we!"'e interviewed; duri:lg
the second visit, 25 were interviewed.
In add:tion to inter'views with members of our primary sample, ap9r'oxi-
mately 3C supple!':1ental interviews wer'e conducted" These included
interviews Hith fir st and secone:! line super'visors and Managers, inter'views
6
with production Horkers in an adjacent depar'tment., and interviews with
ind i vid ual s fr'om eng ineer ing, mai ntenance, per sonnel reI ation s, and other'
plant staff.. Each interview lasted about 10 minutes and contained both
structured response and open-ended questions concerning the robot and the
circLnnstances sur'rounding it.s introduction. He a1 so obser'ved the wor'K-
force during the intr'oduction of the robot and administered a satisfaction
quest ionnaire to pr'od uet ion wor'ker's"
We examined whether our primary sample of employees (i.e", employees
in the department where the r'obot was introduced) at lime 2 differ'ed from
those we interviewed at Time 1 on character'1.stics such as the length of
time employees had workad at the plant, their job grade, and shift. Our
lime 2 sample was drawn without replacement from the population of
employees we inter'Viewad at Time 1; hence our' rime 2 sample was not
independent of our 1ime population. TherefOre, we used the
hypergeometric distribution to test whether our Time 2 sample was
representative of our 'Iime population" We derived from our Time 2
sample maximum likelihood estimates of the frequencies of individuals in
var'ious categories (e"g., first, second, or third shift) in the population
most likely to have generated our' sample. We then comput ed a / test
statistic which compared the probabil ity of drawing our lime 2 sample from
our' I'ime 1 population to the probability of drawing our lime 2 sample from
the population most likely to have gener'ated it. 18 ;>
fhe x- values were not
lar'ge enough to reject at more than moder'ate levels of significance (p <
,25) the hypotheSiS that our Time 2 sample was a random sample dra,m
without r'eplacement frem the population of employees !,.;e int.ervieHed at
Ii~e 1.. rhus lour' sample at Time 2 appears t.o be representative of OUf'
population at Time 1 on these var'iables.
7
R esul ts
rhe results are organized in terms of the effects the r'obot had on
worker beliefs, activities, and inter'actions. Then we examine the effects
of introducing the robot on pr'operties of the or'!lanization.
Beliefs
About the robot. We were curious about how workers in our sample
think about robots, so we asked the respondents an open-ended question:
How would they describe a r'obot to a friend? Table 1 lists the phrases
used to descr'ibe a robot. The rna jor concepts seem to be: mechanical man,
preprogrammed machine, something that loads machines, increases
pr'oductivit.y or reduces manual work. This list of descY'iptions seems to
fall into three categories: gener'al descriptions (mechanical man),
functions (loads machines), and consequences (r'educes manual work). An
examination of the frequency of concepts used in the dHfer'ent classes
during rime and rime 2 shows no significant changes. 19 That is, the
general types of categories used to describe robots r'emain the same.
However', we found a significant increase in the number of concepts
mentioned by each individual over time, t(24) = 1,,89, p < .10. This is
consistent with the idea that more experience should lead to a more
differentiated view of robots.
As a follow up to the fir'st question, we asked worker's how they
learned about r'cbots" rhe movie Star liars and television shows depicting
humanlike robots were frequently mentioned. These humanlike r'obots in the
r:1edia probably contr'ibuted to the tendency we obser'ved at ~he plant fo['
wor kef s to anthr'opcmof'phize the cobot., Hor'kef's on eac~ shift named the
robot and endowed it with human qualities" This tendency was ev ident j.n
both the inte!"'views and in obser'vations of people in the wor'kp12ce"
8
About the ei'f'ects of r'obats" Table 2 {::resents war'kers' beliefs about
robots in gener al at Time 2, Seven questions Of' stat.ements t.Jere r'ead to
t.he respondent, Ttlho then responded by strongly agreeing I ;:lI?;reeing,
slightly agreeing, slightly disagreeing, disagreeing, or' strongly
disagreeing" Ihe results indicate that worker's in our sample had ~sitive
atti tudes toward r'obots" The worker's felt that r'obots '''ill help the
United States remain competitive" There is some indication that war'kers
believed that robots will displace other' wor'kers but not themselves and he
limited to certain types of jobs" Wor'kers perceived that the use of
robots will mean that workers r'equire additional education and skill
training"
Table 3 focuses on workers' per'ceptions of the effects of the r'obot
in their' depaT'tment rather than their general belief's about robots" The
respondents were presented with an outcome (e.g., the chances of an
accident) and asked at Time 1 whether the robot would incr'ease, decrease,
or have no effEct on that outcome. Respondents were asked at Time 2
'"hether' the robot had actually incr'eased. decreased. or' had no effect on
the outcomes" Table 3 pr'esents the information for lime 1 and Time 2
separately in per centages.
accident outcome in Table 3.
For example. consider the chances for an
Ihe re suI ts pr e sen ted in Iabl e , i nd icate
that 11 percent of our respondents at Time thought the r'obot would
increase the chances of an accident; this number had increased to 29
percent at Time 2. rhe number of respondents who thought the robot would
deer'ease t.he chances of an accident was 41 percent. at. Time 1; ~1 percent
of our -:ime 2 respondents thought the robot had decrea sed t.he chances of
an accident.
9
Ihe tesults of the pr'obit analyses 20 «hich tested \.Jhet~er' there «er'e
significant changes in the number' of f'esponses in the increas~, deer'ease,
or no effect categot'ies ftom Time 1 to lime 2 at'e al so ptesented in Table
3. For' example, the tesul ts of the probit analysis on the chances of an
aeci.dent outcome r'eveal that the coefficient of time in the pr'obit model
fat accidents «as -0.687. This indicates that tespondents wer'e more
likel y at Time 2 than at Time 1 to move in the d ir' ec t ion a f sa yi ng the
chances of' an accident increased. This coefficient was large relative to
its standard errot, t(52) = -2.19, P < .05.
Iable 3 indicates that a majority of the «orkers at lime 2 felt that
robots increased productivity, but did not have much effect on the quality
of output, the amount of downtime, or' the number of people who wor'k in the
depar'tment. The results of the probit analyses indicate that wor'ker's were
significantly more likely at Time 2 than at Time 1 to say that the robot
increased the chances of an accident, incr'eased costs, and lower'ed the
quality of the output. This change in per'ceptions ovet time was highly
significant for the quality of the output, t(52) = 3.39, p < .OOL
Tabl e 4 uses the same format but focuses on a different set of
outcomes. The gener'al trend s at Time 2 seem to be that the r'obot
decr'eased the number of boring jobs and teduced fatigue on the job" A
majority of wor"ker's believed that. robots require war'ker's to learn greater
skills, which is consistent with the information in Table 2.
The results in fables 3 and U suggest that worker's in t.he department
Hhere the robot ioTas intr'oduced became less opti:nist.ic over time about t.he
effects of the r'obot" Exa.mining the coefficients of the time variable for'
the eight outcomes on which we have data for both Time 1 and 'Ii:ne 2
reveal s that ;'es;1ondents ' .... er'e less optimistic about the effects of :'he
10
robot on six of these eight outcomes at Time 2 than they were at Time 1,
Ihe coefficients foF' thr'ee of t.hese outcomes (accidents, costs I and
quality) were significant" I'he only exceptions to this trend toward
greater pessimism are the nonsignificant coefficients for' the number of
people who work in the depar tment and the number of jobs that ar'e highl Y
skilled.
These questions wer'e also asked of pr'oduction workers in a depar'tment
adj acent to the one in which the robot was introduced in order to
understand how per ceptions of workers in other depar'tments in the plant
were affected by the introduction of the robot" Harkers in this adjacent
depar'tment had access to some of the same sources of infor'mation about the
robot (e "g., the demonstration, the plant manager's talk, and the notices
discussed above) as did workers in the department with the robot" '"orkers
in the adj acent department were al so able to watch the r'Obot operate. The
patter'n of results obtained ftom worker s in the ad jacent depar'tment was
similar to that descr'ibed previously for' our primary sample,,21 Ihere was
some evidence that worker's in the adjacent depar'tment were more optimistic
about the effects of the robot at lime 2 than were their counterparts in
the department with the robot. For example, workers in both departments
wer e more li kel y at Time 2 than at lime 1 to say that the chances for' an
3ccident increased and t.hat costs incr'eased; however', workers in the
adjacent unit thought that the chances for an accident and cost.s increased
t.o a smaller extent than did workers in the depar'tment with the robot.
About introducing change. In or'der to understand how workers
actually J.ear~ed about the robot, l,.;e asked the employees at I'ir.1e a
se!' ies of questions about whethe!" t.hey lear'ned about t.he tahot f'r em (3
par'ticuler sour'ce and about the extent to which the source increased their
11
understanding of' the robot" Ihese data, presented in Table S, indicate
that the most frequently mentioned sour'ce of infor'matioo about r'obots Has
the weekly wor'kplace meeting betHeen supervisors and T,-IOr'ker's .. POHever I
these meetings increased workers' understanding of r'obots only to a little
extent" HY'itten communication and the demonstration at the open house
were the most effective sources of information about the robot. HOHever' ,
less than half of our respondents attended the open house and only 16
percent reported that they received a written communication" Thus, t.he
var ious comrnunicat. ion sour'ces do not seem to have been very helpful in
increasing Hor kers' understanding of the robot.
In addition, we asked wor'kers how much influence or involvement they
actually had on decisions about' ( 1) whether the robot would be
introduced in their department, (2) where it would be placed, and (1) who
would run it. Also we asked them how much influence they should have had.
rhe wor'kers reported they had no influence on any of these three
decisions. Ihey said they should have had a little influence on decisions
about whether the robot would be introduced and who should run it"
Harkers did not think they should be involved in decisions about where the
robot was placed.
Activities
Our analysis thus rar has focused on beliefs. ~!o W i.re t.ur n to the
quest,ion of how the robot introduction affected workers' activit.ies. Our
sample now is com~sed of the individual on each shift who operated t.he
r'o bot" Speciel interview schedules were devel·:)ped to !"'1easure the
act::'vities. war k cycle, and interaction patter'ns of the operator's' johs
before ar.d after the tobot was intr'oduced.
12
j'lhen the robot was introduced, a manufacturing cell Ci"e", a set of
interdependent machines oper'ated by a worker) was cr'eated" The robot
provided material handling functions for' two millin~ machines" An
operator was then responsible for' the two milling machines and the r'oboL
The introduction of the robot removed the materials handling activity from
the oper'ator,1 s job and added a new activity, r'cbat oper'ator' ..
Prior to the robot's introduction approximately twelve different
products were passed through both milling machines to begin the work flow
sequence for thi s prod uction depar'tment. The T..Jork cycle for' each machine
included set up activities and then relatively short milling times (1-2
minutes) during a pr'oduct run. Work was done to fine tolerances and it
was possible for the operator to determine, thr'ough sane measur ement
procedures, whether the parts were milled correctl y" The major' quality
control activity, however, was at the end of all the machine operations.
After the change, the number of products remained the same. The two
milling machines were still located at the beginning of the work cycle
where the bar stock went through the first two milling oper'ations. The
other machine operations were unchanged" Quality control activities were
still at the end of all the machine operations.
Ihe major change in aotivities in the new manufacturing cell came in
the material handling activit.ies. In the old system, the operator would
pick up t.he stock, place it on machine one, clamp it in, star't the
:nachine I and then the milling machine would perform it.s wor'k. Then the
operator ''';Quld stop the machine, unclamp the stock, place it on the second
:":12chir:e and t.he cycle would repeat" Each ne1d stock would folloH this
cycle" Uhen we asked the oper'ators about the differences bet'deen their'
jobs beior'e and after' the r'obat intr'oduction, they said:
13
Now it's mainly watching &
sure everything is running., " walking arour.d the m8chines to be
He do mor's activities. machines.
Now you have to set up all three
There are also more functions robot.
. you need to progr am the
So we see a shift from manual to cognitive activites (monitoring).
Al eo, the oper'ators reported that they were doing more activities and that
the total wor k cycl e had incr'eased. The increased wor'K cycle was
attributed by the operator's to more set ups and delays in getting the new
robot operational.
Ihe change in activities was related to a ch~nge in skill
requirements" The operators said:
Ihe job now requires mor'e skills program the robot and run it. comes more responsibility"
. " . You have to learn how to • With more skills of course
Oper'ating the robot requires more skills sophisticated.
. the job is more
If we combine the ideas in these quotations with those in the preceding
ones, it is clear that the new skills appear in the area of observing and
detecting problems in the inter'face between the three machines, and in
progr'amming and operating the robot.
What are some of the consequences for the worker that result from
these changes in activities? The general literature on the relationship
bet1Neen job activities and individual characteristics indicates that
improving the fit between the new job activities and the personal
charseteristies of the war'ker can lead to more positive at.titudes and
higher ::J.otivat ion" Conver'sely, intr'oducing activities that ar'8
incompatible ylith a war'ker" s abilit.ies and preferences is likely to
gener'ate str'ess, negat.ive attitudes. and lower motivat.ion. Given :hat T .. .re
14
are examining the introduct.ion of a single f'obot and are dealing with only
three oper'ators, j,t is difficult to identify statistically si.gnificant
changes in our' study" However', some trends are evident." The oper'ators in
our study ex perienced mar'e str'ess or pressure" Interview comments from
two of the operators were:
Ihere is more stress now" They want the robot to That's hard because it
We have more responsibility., r'un and we have to keep it going;., is still relatively new"
It's nerve r'acking " there are lots of details ., it's an expensive piece of equipment.
This stress stems partly from the new tasks and responsibilities of
the oper'ator's, and par'tly fr'cm oper'ating a net,,, and costly piece of
equi pment . Iher'e was another more subtle source of stress which arose
from worker s' comparing themselves to the robot. There was much
speculation during our fir st visit to the plant about whether an operator
who was par t icularly quick would be able to beat the robot. By our' second
visit, workers seemed r'esigned to the fact that the r'obot would always be
able to outproduce a human worker. The reason was simple: Robots do not
take breaks or' even go to lunch!! Objectively, after the robot
intr'oduction, the operators contr'oUed the robot and the two milling
machines. However operators subjectively viewed the situation as one of
cc:mpetit ion between them and the machine. Operators reported that the
robot could load and unload the two milling machines faster than the
oper'ator's could when they oper'ated the machines manually" Ihi s may be
another source of stress.
:-ne of the robot. oper'ators in our sample mentioned that although r:e
found observing and monitoring more boring than manual act.ivities, he :,.J8S
cU!"Tently satisfied with his job because there were still many of 1YJanual
set-up activities. The oper'ator COmm8!1ted, hOHever that vii thout these
15
set-up activities, the CUYTent job would be more boring; than the previous
job. ~Je think this incompatibility between activities required by the job
and pr eferences of t.he worker was another sour'ce of stress.
Now, the picture we want to dr'gw is not one of an unhappy operator.
All these individuals voluntarily accepted the job of operating the "lanu-
facturing cell. They all received considerable recognition because of the
"newness" of the robot., All operators acknowledged that the prior heavy
and fatiguing wor'k was eliminated by the robot. The r'Obot operators
reported approximately the same degree of satisfaction with most aspects
of their' job before and after the robot was intr'oduced.
At the same time, worker's experienced mar'e stress than they had
experienced in their' pr ior job. We have attempted to identify potential
so Uf'ces of this s tre 55. It would be premature at this point to speculate
whether this increased stress was good or bad for the individual or the
organi zation" Studies have shown that. incr'eased stY'ess is associated with
increased turnover and absenteeism22 and that stress can lead to either'
increments or decr'ements in per'formance. 23
Since the introduction of the robot could affect support personnel
activities as well as operator's, we interviewed people in engineering t
maintenance, quality control, and scheduling" Engineering and maintenance
were responsible for getting the machine up and running. Changes in the
functioning of the machine could affect quality control and scheduling"
Sever a1 themes emerged from our' discussions with suppor't per sonnel '.
F'irst.. some support personnel felt the r'obot had changed their job
activit.ies .. Si nce t.he robot r'e pr e sen ted a new gener at ion of t ec hno j. ogy .
neH knor,.,rledge and new job activities Her'e necessar'y" Second. there wer'€
some feelings of frustration since not all of the suppor't per sonnel Her'e
16
involved in planning the int.roduction of t.he robot, yet t.hey i,Jere expected
to acquire new \<nowledge and skills" Third, there wer'e poslti'.Je feelings
associated with being r'ecognized and per'sonal pride derived fr'an the
successful oper'ation of the robot. And fbur'th. the magnitude of changes
in activities and feelings 1,,[8S T'elatively modest. given that we ar'e
dealing with only one robot installation.
Inter action
Of impor'tance in the wor k place are the formal and informal inter'-
actions that develop around job activities" The introduction of a r'obot
can change these inter'action patterns, which in tUT'n can have
psychological and behavioral consequences. For ex am pIe. if' the new
technology breaks up eXisting social interactions and isolates the worker.
we expect incr'eases i,n alienation and mor'e resistance to the nefti
technology"
Ihe robot operators reported at Time 2 that they had less opportunity
to talk with people on the job than they had before the r'obot was
introd uced . fwo of the operator s said:
I haven't been able to talk as much. I'm too involved with the robot. You really have to concentrate.
I don't have time to talk wi,th anyone. them breaking my concentration"
I don't want I'm isolated now.
The decreased opportunity to inter'act with others seemed to derive mainly
from the increased mental demand s of the job. l,vor'kers had to concentrate
more., They did not have time to talk with coworker's.
Ihe introduction of the robot did not change the work flow in the
department. All the iolorkers, including the r'obot operator', wer e loca+:,ed
in the saMe aT'ea and par'ticipated in the same part of' the \.;or'K f'IOH"
17
Thus, while our' operator's reported less opportunity to inter'act !.-lith
others in the department. the set of people they interacted '-lith within
the depar'tment remained r'oughly the same. Ihi s might have pr ov id ed
buil t-in support mechanisms to buf fer the worker s from some of the effects
of the aha nge .
The major changes in interactions occur'red between support. personnel
from engineer ing and maintenance departments and the oper ator's of the
manufac turing cell. There was more fr'equent contact among engineering,
maintenance. and the robot operator" Perhaps because the robot was new
and repr'esented the first major installation in this factor'y. support
personnel as Hell as the robot oper'ators were highly motivated to get the
robot up and running and cooperated with each other. If there were many
robots being placed on line and the suppor't personnel had great demands on
their time. then we might find more conflict between suppor't per'sonnel and
oper ators.
Organizational Unit
Our discussion thus far has focused exclusively on the effects of the
robot on the individual.
affected the depar tment.
lhe introduction of this new technology a1 so
Spec ificall y. the introd uction of the robot
requir'eG a r'e-evaluation and reclassification of the oper'ator" s job"
Because cer'tain job activities were e;Liminated and other' activities wer'8
added. the quest ion was whether the net change indicated that the job
should retai:1 the same grade or be upgraded. Hanagement did upgrade the
job. out worker's felt that the new grade and associated pay for the
oper'at.or's job were still too low,.
He also examined Itrhether the intr'oduction of the robot affected other'
depar'tment pol ici es, proced ur'8S, or formal coor'dinat ion mechani sms" fhere
was no evidence of any effect other than the changes in the pay system.
Discussion
The purpose of this paper is to focus our attention on the
consequences for' the worker of intr'oducing robots into the factory"
Workers in our sample held positive beliefs about robots in general. When
we asked them about the effect of the r'obot on their department. they
ini tiall y repor ted that the robot would increase productivity. reduce
costs, increase quality, make the job easier and less bor'ing, and increase
skill requirements. As workers acquired mor'e experience with the r'obot
their beliefs about robots became more complex and somewhat more
pessimistic (greater chances for' accidents, cost increases, quality
decr ease s)
We examined how the introduction of the T'Obot changed the production
operators' work activities. The operators' work activities shifted from
primarily manual activites (lifting) to cognitive activites (monitor'inil)"
The operators reported that they "Iere performing more activities and had
more responsibility. The changes in activities were related to feelings
of more stress" Of course I ther'e were offsetting benefits as the most
tiring aspects of' their' jobs had been eliminated and the robot operators
received greater r'ecognit.ion and r'emuneration as the first robot was put
on line"
per sonnel ,
Changes in activities also appeared in the jobs of SUODort
Again, there was some evidence of stress created by the ne\ol
;,mr'k acti."ities,
'9
Interaction patter'ns changed for the robot operators .. Operators
reported that they felt isolated and did not have as much time to interact
t,..;ith cowor'kers in their' depar'tment. Ther e was mor'e fr'equent contact
between individuals fr'om engineering and maintenance and the r'obot
operator's. However, the modal job and social interaction patterns for the
department before and after the robot introduction remained essentially
the same.
Ihe introduction of the robot had an impact at the organizational
unit level, particularly in the area of job evaluation and pay" Ihere was
a widely-held belief on the part of the workers that the robot operator
job had been unfairly evaluated and the pay grade was too low.
The process of introducing technological change was also examined.
Ihe basic finding was that there was a discrepancy between management's
attempt to communicate about r'Obots and the workers' need to learn about
robots" Ihat is, although many communication techniques were used, few
were received by the workers. 1hose communications received by war'kef's
were not seen as particular'ly helpful in increasing their understanding of
the r'obot.
We think this study has several major strengths" Io our knowled~e,
it is the first systematic evaluation of the effect of introducing a robot
on workers" Ihe study used a variety of methods, includinQ; interviews I
questionnaires, and observations. lile collected data before and after the
change to get some baseline to study the effect.s of the intr'oduction over
time" In addition, ',.;e used a broad sample per'spective to identify people
dir'ectly affected by the change as well as those indirectly af'fecte~
(e"g .. , support per'sonnel),
20
Ihe study. of CQUf'Se, has certain limitations" Dat.a -"'ef'e collect.ed
fran only one organization. 1here was only one r'cbat installation and it
was the first T'Obot installation .. 1he change al so teok place in a non-·
union organization where ther'e ~.,ere positive ('elations betHeen labor' and
management. While intervieTd data on the effects of the robot on various
outcomes such as pr'oductivity were collected. company records data on
these outcomes wer'e not collected. F'uture resear'ch involving mul tiple
of'ganizations and including f'ecords or archival data as well as interview
data collected over several points in time (e.g., before. shor'tly after,
and a year or so after' the robot introduction) is needed.
What can we learn from this study that will help in new installations
of f'obotics in factories? Despite the small sample size and the
difficulty of systematically testing certain r'elationships, a number of
findings have emerged from this study. These findings combined with
f'lndings from other studies of increased automation suggest some possible
recommendations for manager's introducing new technologies ..
1) Prior to any intr'octuction some questions need to be r'esolved.
Questions concerning job security and pay are likely to be uppermost in
the minds of the work force. Failure to resolve these questions prior to
the introduction of the robot is likely to reduce the effectiveness of the
introd uction .
2) Diagnosis of the organization prior to introducin~ change is
critical" h'hat effects 'Nill the technological change have on activit.ies,
interactions, and belief's of 'Nor'kef's? Pr'oblems caused by the change need
to be anticipated--sane will be obvious (in cases of job loss) Hhile
other's ~ill be subtle (in cases of new job activities)
2 1
3) A. strategy for' war'ker' involvement in introducin!S this new
technology needs to be delineated" There ar'8 a wide var'iety of possible
strategies" Management in our study pr'ovided virtually no oppor'tunities
for' Hor'ker involvement in this technological change. 1he Hor'kers wanted
some levels of involvement for' cer tain decisions but not other's" ,Some
levels of involvement are likely to increase understanding about the r'obot
and per haps lead to greater commitment to the change pr'ocess.
4) Certain communication techniques seem more effective than others
in intr'oducing robots. Demonstrations that illustrate the operations of a
robot seen to be powerful techniques"
5) Some feedback mechanism to monitor' communicat.ion effectiveness is
necessary in introducing this technology. Our study showed a di screpancy
between what management was trying to communicate to workers and what the
workers received"
6) It is vital that fir'st line supervisors be given information about
the robot and support from upper management in dealing with Harkers'
reactions to the r'obot. In times of change workers are likely to go to
their' supervisors mor'e frequently for information and advice. The
attitudes and behaviors of supervisors are likely to have a big effect on
the success of the robot introduction"
n The robot will create neH job activities. It is very important to
do a car e ful anal ysis of the new job and max imi ze the fit between job
char'3cteristics and personal characteristics. The literature on
job-per son fit indicates that a lack of congruency may have dysfunctional
effects on the person and organization. Ihe quest ion is not ~.mether the
worker" ca!'] do the !'leW activities but whether' the worker can do and pr'efers
22
these activities" If ther'e is lack of congr'uency, one must consirler
alternatives in job r'edesign or in selection procedures,
8) If the change is frem "doing" to "observing" activities, the
worker s may ex perience more boredom in the job. If t.his occurs, some
mechanism to alleviate boredom, such as job rotation, may be helpfuL The
job rotation would increase task variety and build up a backlo~ of skills
for' future expansion of r'obotics.
9) Training backup operators for the robot is impor'tant. In our
study only one per son per shift was initially trained to operate the
robot, Ihis l~d to disr'uptions in the 1,olor'k pr'ocess when one of t.he
operators was absent.. Training backup oper'atoT's would pr'ovide the
organization with more flexibility, and individual workers with more job
variety.
10) The introd uction of tobots can affect the nature of social
interaction patterns at work. Prior' resear'ch shows that attempts to
change these patter'ns can gener'ate r'esistance to change .. If diagnosis
indicates that the change will break up eXisting social relationships,
some a1 ter'native strategies need to be conceived. For' example, involving
the worker in this par't of the change may generate new work arrangements
that wi] 1 facil Hate the acceptance of change.
11) A successful introduction of robots requires the cooper'ation of
the support personnel. Our study showed that not all of the support
personnel He!"8 involved in planning t.he intr'oduction of the robot, and
SOMe stress \Vas c?"eatect as a r'8sult of the lack of part.icipat.ion.
::::1volvement of the suppor't per'sonnel and oper'ator's early in the changA
process should facilit.ate the introduction process ..
23
1 able 1
',VORKERS' DESCRIPTTONS OF ROR01S
Mechanical Man
Hydraulic Arm
Comput.er
Pr epr ogrammed Machine
I.. V. Image
Hoves Material
Loads Machine
Better Productivity
Reduces Manual Work
'darks Cant inuousl y
Percent of Total Mentions
Time 1 Time 2
1S% 9%
2% 9%
6% i)«,
15% 1M
10% 7'1,
4% Q,%
12% 14<>1,
15% 5%
15% 2~%
6% A%
100% 100~
24
lable 2
HaRKERS' BELIEFS ABOUI ROBOTS IN GENERAL TH'E THO
Robo ts will:
Make the U"S. more competitive
Be ca pable of doing my job
Percent Worker's Agreeing/ Strongly Agreeing
29%
Be capable of doing clerical jobs
Be capable of doing management jobs 17~
Displace wor ker' s SO%
Create less desirable jobs 21~
Require mor e job r'etr'aining
Table ~
PERCEPTIONS OF THE EFFECT OF A ROllOT IN A PROOIICTTON OEPARTMENT
a Percent of Responnents Who Replied That the Robot Had a Certain Effect on:
Prodllctlvlty ACCIdents Costs QuaU ty Oowntlme ~nher:-".~~!eople
Effed: Time 1 Time 2 Time 1 Time 2 Time 1 Time 2 TIme 1 Time? Time 11> Time? Time 1 Time:.? ------Increa!1e 81 6'7 11 29 10 33 ')1 11 iA I) o
No Effect 8 12 12 ~6 7 II 10 ')11 ')11 1111 Al
Decrease 3 12 111 21 55 ~2 1 21 A 'III 17
Coefficient of Time VarIable UI Pr'obi,t t10del
c 0.607 -0.687 -0.760 1. 1/)7 -0.'>i1
Standard Error of l'lme Coefficient 0.392 0.31'1 o .~O'I o . i'III (). "1?')
T-~)tatlsti.c 1.5118 -2. lila -1.879 3.19'1 -1./)3~
~)j grll fi.canee 0.128 o. 03~l 0.06A 0.001 0.ln1
a The fX!fCent of re!lPondents who replied. "Don'L Know," 15 not included in Uns table; hence, percent,Flges for sorne (liltcomes will sum to less than 100.
b R(!~p0l'ldents Here not asi(ed about thIS outcome at Time 1.
e A posiLlve eoefficlent Indi.cates that respondents were more likely at Time 2 than at Time 1 to move in the dlreeLtOIi of :JaYlng the particular outcome decreased; a negative coefficient. indicates that respondents were more likely ,It Tillie? than at 1'11' 1 to m~v' in the directton of saying the outcome increased.
Table II
PERCEPTIONS OF HIE EFFECT OF A ROBOT TN A PRODIICTION DEPARTMENT - -
a Percent of Respondents Who Replied That the Robot Ilad a Certain Effect on:
Skill Chances for a !!.~ring Jobs Requirements Better Job Fatigue Morale Absenteeism
E ffecl; Time 1 Time 2 Time 1 Time 2 Time 1 Time 2 Time lb Time ;> b
Time 1 Time;> T " h 1'" Ime 1 Ime? -----
Trlerease 16 20 511 6~ 21 17 II 2S 12
No Effect 19 16 3~ 36 59 75 R 7"> C) II
Decrease 59 60 0 0 9 R 19 o 12
Coefficlent of Time Variable HI Pr'obiL Modele -0.052 -0.012 O. lOR
Standard Error of Time CoefficIent 0.321 o. 311~ 0.326
T-StC.ltistic -0.163 -0.208 0.313
Signlfieance 0.871 0.816 0.141
-"------"-~-----
a Tile p{:rcent of respondents who replied, UDonlt Know," IS not 1ncluderl In this table: hence. percentaR~S for some outeomes wi 11 StUO to less than 100 ..
b Hesl)ondenLs were not asked about this outcome at Time 1.
C A positIve coefficient Indicates that respondents were more like] y at Time 2 than at Time 1 to move in the (Hr'ectlOIl of 5aylng the partIcular outcome decreased; a negative coefficient indicates that respondents \WI~e more likely at Tillie 2 than at Time 1 to move 10 the direction of saying the outcome increAsed.
27
lable 5
EFFECIIVENESS OF CO~'~'lJNICATION
Communi,cation Sour ce
lt1r'itten communication
Hor kplace meetings
Communicat.ion fr om super-v isor
Movies or audio-visual presentations
Demonstr ations
Informal sources including the grapevine
ABOU1 THE ROBOT:
rIME ONE
% Workers Reporting that Ihey
Received Communiction
16%
89%
46%
13%
42%
37%
Average Extent Communication Increased Horkers'
Under' stand i ng*
2.6
4
4. 1
2.7
4
• Ihe response alternatives were: (1) to a very great extent, (2) to a great extent, (3) to a fair extent, (4) to a little extent, COi) not at all.
28
REFERENCES
1 See "Robots Join the Labor' Force ,II B1Jsiness Heek, 9 June 1C)RO.
2See A .. Salpukas, "Manufacturing Using Robots," The New York limes, ~~ October 1980 ..
3J .. Engelberger, Robotics in Pr'actice, (New yor'1<: Amacom, 19An).
4 R .. U. Ayres, "Ihe Futur'e of Robotics: A Meta Review," Prepar'ed for The UNESCO/ll ASA Symposium on Scientific For'ecasting and Human Needs: Methods Tr'ends and Message, Ibilisi, USSR, December 1981.
5R. U. Ayres and S. M .. Miller, "Preparing for the Growth of Industrial Robots ," Unpublished manuscript, Carnegie-Nellon University, 1981..
6Eikonix Corporation, "Technology Assessment: The Impac t of Robotics," Iechnical Report EC/2405801 - FR-l, Eikonix Cor por'ation , 1979,
7F • E. Emery and E, l. Trist, "Socio-technical Systems," In F. Baker to Complex (Ed .. ), Organizational Systems: General Systems Approaches
Organizations (Homewood, II: Richard D .. Irwin, 1973).
8W• F, Whyte, Men at Work (Homewood, Ii: Richard D. Irwin, 1961).
9D • Elizur, Adapting to Innovation (Jerusalem: Jerusalem Academic Press, 1970 ),
10 F , C. Mann Holt, 1960),
and L. R. Hoffman, Automation and the "arker
ll E• Mumford and D. Banks, The Computer and the Clerk (London: and Kegan Paul, 1967).
(New Yor'k:
Routledge
12". F. Whyte, Nen at Work (Homewood, TL: Richard D. Ir win, 191) 1).
13L., K. Williams and C. B. Hilli""s, "The Impact of Numerically Contt'olled EqUipment on Factor'y Organization ," Califor'nia Management Review, ltlinter' 1964, pp" 25-34.
14tL Beer, Or'ganization Change and Development (Santa ~'lonica, CA: Good year, 1980).
15 L. each and J. R .. p" French, "Overcoming Resistance to Change," Human Relations, Vol. 1, August 1948, pp. 512-532.
16 L £" Griener. "Patterns of Organization Change," P'aT'var'd Business Review, Vol" 45, No.3, May-June 1967. pp. 119-128.,
1\;. Crockett, !1Introducing Change to a Government ftgency," Tn P. H" Hirvis ~nd D. N. Berg (Eds.) I Failures in Organization Development and Change (NeT"'; YOf'k: tiiley, '1977).
29
18Details on the use of this analytic procedure as well as the r'esult.s of these and subsequent analyses are available on request from the first author', Thanks are due Dennis Epple and Lars Hansen for calli,TIg; our' attention to these issues.
19 To test T..mether the f'r'eqeuncy of mentions in various categoT'ies changed from lime 1 to Time 2, we used a similar pr'ocedure to that outlined earlier for testing whether our sample of employees differed from lime 1
to TIme 2 on key characteristics" In part.icular, we pooled the freqeuncy of mentions in each category at Time 1 and Time 2 to arrive at the frequency of mentions in the population. USing the hypergeometric distribution, we then tested whether the distribution of mentions at Times 1 and 2 were randcrn samples drawn without replacement from our population. Ihere was no evidence for r'ejecting the hypothesis thjlt our' Time 1 distribution was a randan sample fran the population, x-(2) = 1.04, p < .75, or for r'ejecting the hypothesis ttpt our Time 2 distribution was a randan sample fran the population, x-CO) = 0.82, p < .75.
20rhe N-Chotor.lOUS probit model requires independence of t.he error t.erms. This requirement might be violated if, for example t observations for the same individual at different points in time were correlated" In using theprobit model, we have assumed that this dependence, if it exists, is weak.
21 Ihe pr'obit analyses on the outcomes discussed above were r'un separately for each department with time (Time 1 vs .. Time 2) as the predictor' of the outcomes. lhe probit analyses were also performed on the data from both departments combined, with both time and department (department with the robot, adjacent depar'tment) as predictors of the outcomes.. A similar pattern of results was obtained fr'an the differ'ent analyses.
22L W. Por ter and R. factor's in employee 1973, 80 151-176.
M. Steer's, turnover and
"Or'ganizational, work, and per'sonal absenteeism, If Psychological Bulletin t
23~-l. E .. Scott, lfActivation theor'y and task design ,11 Organizational Behavior and Human Performance, 1977, 111-128.