The Sharing Game: Fairness in resource allocation as a function of incentive, gwndwr, and recipient types

Judgment and Decision Making, Vol. 2, No. 3, June 2007, pp. 204–216
The Sharing Game: Fairness in resource allocation as a function of
incentive, gender, and recipient types
Arthur Kennelly? and Edmund Fantino
University of California, San Diego
Abstract
Economic games involving allocation of resources have been a useful tool for the study of decision making for
both psychologists and economists. In two experiments involving a repeated-trials game over twenty opportunities,
undergraduates made choices to distribute resources between themselves and an unseen, passive other either optimally
(for themselves) but non-competitively, equally but non-optimally, or least optimally but competitively. Surprisingly,
whether participants were told that the anonymous other was another student or a computer did not matter. Using such
terms as “game” and “player” in the course of the session was associated with an increased frequency of competitive
behavior. Males were more optimal than females: a gender-by-incentive interaction was found in the ?rst experiment.
In agreement with prior research, participants whose resources were backed by monetary incentive acted the most
optimally. Overall, equality was the modal strategy employed, although it is clear that motivational context affects the
allocation of resources.
Keywords: distributive fairness, gender, human-computer interaction, monetary incentive, resource allocation, Sharing
Game.
1 Introduction
son had calculated that while such a strategy would not
yield maximal amounts for either party, it would reduce
Imagine being repeatedly given the choice between re-
any discrepancy in earnings between them. A third noted
ceiving $5 and $7. The obvious choice is to take the $7
that his decisions may differ depending on how the situa-
every time. Now imagine that this choice comes with the
tion was framed. The intent of this paper is to investigate
following strings attached: if you select the $7 for your-
the patterns of distributional choices made by people in a
self, then an anonymous other will receive $9, but if you
simple economic situation such as that described above,
take the $5, that anonymous other will instead receive $3.
and to assess if (and to what extent) certain contextual
While consistently choosing the $7/$9 option is still the
variations affect these decisions.
optimal choice (by “optimal,” we mean the choice that
Popular economic games examine how participants al-
yields the maximum amount for the chooser), might it be
locate resources.
For example, two commonly stud-
bothersome to know that this unknown person is receiv-
ied games are the Ultimatum Game (UG; Güth, Schmit-
ing more than you for doing nothing? When this repeated
tberger, & Schwarze, 1982) and the Dictator Game (DG;
choice was posed to members of our lab, not all of them
Forsythe, Horowitz, Savin, & Sefton, 1994). In the UG
elected to take the optimal path. One of our colleagues
one student proposes a distribution of resources (for ex-
insisted that he would select the $5/$3 option every time,
ample, if $10, $6 for him and $4 for the other player). If
stating, “I’d want to make sure that I have more than the
the other player accepts, the $6 / $4 split becomes reality.
other guy.” To this person, it was well worth it to sac-
If he rejects the offer neither gets anything (no negotia-
ri?ce a couple of dollars to ensure having a relative ad-
tion is possible). In the DG whatever the proposing par-
vantage over the stranger. Another stated that she would
ticipant proposes becomes reality (the second “player” is
take the unusual path of alternating between the options
passive). It is also instructive to ask participants to choose
from trial to trial: “I’d get $7 and then $5, while he’d
between two possible ?xed allocations of resources be-
get $9 and then $3, so we’d both end up with the same
tween themselves and another player (e.g., Bazerman,
amount — $12 each — after every other trial.” This per-
Loewenstein, & White, 1992; Falk & Fischbacher, 2000).
We report two such studies assessing college students’
?Research supported by NIMH Grant MH57127 to the University
allocation of resources to themselves and another player
of California, San Diego. Corresponding Author: Arthur Kennelly, De-
partment of Psychology 0109, University of California, San Diego La
where the allocations involve points either with or with-
Jolla, CA 92093–0109. E-mail: kennelly@ucsd.edu
out monetary value.
204

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
205
Half the participants are told that the other player
personality characteristics. We ask how choice is affected
is a person and half that the other player is a com-
by monetary incentive and by other central aspects of the
puter. According to the “Computers Are Social Actors”
game (such as gender of participants, whether the other
or “CASA” model (Nass, Steuer, & Tauber, 1994; Nass,
player is designated as being another person or a com-
Steuer, Tauber, & Reeder, 1993; Reeves & Nass, 1996),
puter, and whether the competitive aspects of the game
the social rules applying to human-human interaction ap-
are made more salient).
ply equally to human-computer interaction, implying that
In addition, the Sharing Game paradigm pits compet-
participants in the Sharing Game might treat computer
ing predictions from two leading types of social prefer-
players the same way as “human” players. It was uncer-
ence theories against each other.1 Theories of inequality
tain if such equality would be seen in our experiments
aversion (Fehr & Schmidt, 1999; Bolton & Ockenfels,
though, due to the fact that participants in the Sharing
2000) state that in economic situations, people tend to act
Game do not directly interact with the recipient; rather,
to minimize the difference between their own and oth-
their choices act upon the other in much the same way
ers’ payoffs. These theories would predict Sharing Game
that proposers’ decisions act upon recipients in DG. In
allocators to choose so as to reduce or eliminate the dif-
no way does the recipient act upon the allocator in either
ference between players’ totals (which can be ideally ac-
game. In contrast, the participants and computers in the
complished by alternating between the payoff choices in
CASA experiments always performed some sort of two-
each pair of trials). On the other hand, Charness and Ra-
way interaction with each other, whether the participants
bin’s (2002) theory of reciprocal fairness would predict
were told that the computer was running a program or
that players would consistently choose the optimal option
was acting as the medium through which another person
so as to maximize social welfare.
was communicating. We predicted that such lack of “per-
The issue of whether or not ?nancial incentives af-
sonality” on the part of the computer in our experiments
fect decisions in economic games has been discussed ex-
would in?uence our participants to choose optimally (i.e.,
tensively. Hertwig and Ortmann (2001) noted that the
to maximize one’s earnings regardless of how much or
use of real ?nancial incentives as opposed to hypothet-
how little the other gains in the process) more often when
ical ?nancial incentives often distinguished between the
paired with a computer recipient.
research of economists and psychologists, respectively:
For half the participants the points earned have mon-
“Economists generally pay participants on the basis of
etary value; for the others they do not. This manip-
clearly de?ned performance criteria; psychologists usu-
ulation assesses a question, sometimes contentious be-
ally pay a ?at fee or grant a ?xed amount of course credit”
tween economists and psychologists, of whether compa-
(Hertwig & Ortmann, 2001, p. 383). Fantino and Stolarz-
rable results can be obtained with and without monetary
Fantino (2001) noted that several experiments from their
incentive (Camerer & Hogarth 1999; Fantino & Stolarz-
laboratory failed to ?nd performance differences as a
Fantino, 2001; Hertwig & Ortmann, 2001). Finally, al-
function of real vs. hypothetical incentives (e.g., Goodie
though all of our participants see their own cumulative
& Fantino, 1995, in a study of base-rate neglect and Case
scores throughout the game, half of the participants in
& Fantino, 1989, in a study of the reinforcing effec-
one of our experiments see the other player’s cumulative
tiveness of information). Subsequently, Stolarz-Fantino,
score as well, a factor that might increase the competi-
Fantino, Zizzo, and Wen (2003) found no effect of imme-
tive ?avor of the task and therefore increase choice of the
diate ?nancial incentives on performance in a conjunc-
smaller outcome. Messick and McClintock (1968) previ-
tion fallacy task. Camerer and Hogarth (1999) report
ously found there to be no signi?cant difference between
that performance in certain types of economic game ex-
these two display conditions, but their results involved
periments is helped by ?nancial incentives while perfor-
a game in which both participants had equal decision-
mance in other types of economic games are not. The
making power.
Sharing Game bears a similarity to the dictator game in
Although van Lange, De Bruin, Otten, and Joire-
that one participant is the sole decision maker with re-
man (1997) assert that people generally exhibit stable
spect to how resources are to be allocated. In their survey
preference patterns, De Dreu and McCusker (1997) and
of dictator game experiments that used hypothetical and
Fantino and colleagues have argued that contextual vari-
varying amounts of real money as reinforcers, Camerer
ables can and do affect persons’ behavior in choice situa-
and Hogarth (1999, p. 24) reported that
tions (Fantino, 2001; Fantino & Stolarz-Fantino, 2003a).
The present study asks whether the distribution of the
subjects usually kept substantially more when
Sharing Game strategies is affected by the economic con-
choices were real rather than hypothetical. . .
text in which the game is played. A corollary of this ques-
1The Sharing Game does not directly address a third leading cat-
tion concerns the extent to which the strategies are rela-
egory of social preference theories known as reciprocity (e.g., Rabin,
tively stable as they might be if re?ective of fundamental
1993; Falk & Fischbacher, 2000).

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
206
we can interpret subjects as having some non-
and the other player? If not, how would the frequency
?nancial goal — to appear. . .
generous. . .
of non-optimal choices (choosing the smaller payoff)
— which is partially displaced by pro?t-
be affected by each of the three variables (nature of
maximization when incentives are increased.
other player, monetary incentive, and display of the other
This kind of incentive effect is fundamentally
player’s cumulative score)? Would participants who do
different from the effect of incentives in inspir-
not consistently choose optimally consistently choose the
ing greater effort, clearer thinking, and better
“competitive” outcome (lower payoffs, with the partici-
performance.
pant receiving the larger share)? Or would they equalize
the payoffs for the two players?
If our Sharing Game participants have a similar goal
of wishing to appear generous, then we can expect them
to behave competitively less often when points are not
2 Experiment 1
backed by real money. In addition, Hertwig and Ortmann
(2001) call for “learning more about the speci?c condi-
In this experiment the effects of the three variables dis-
tions under which payoffs improve, do not matter to, or
cussed above were investigated in the context of instruc-
impair task performance.” From a theoretical perspec-
tions which had a competitive (game-oriented) ?avor.
tive it appears important to further clarify the conditions
Within this instructional context, we were interested in
under which ?nancial incentives affect performance, as
assessing the role of monetary (versus non-monetary) in-
Hertwig and Ortmann (2001) have argued. It is also im-
centive, the role of the nature of the other player (person
portant from a pragmatic standpoint: if certain studies
or computer), and the role of presentation (or not) of the
of economic games produce the same results with hypo-
other player’s cumulative score.
thetical and real incentives, then a great deal of human
participant money may be saved.
2.1 Method
The Sharing Game task used in the present experi-
ments shares a characteristic with the popular Prisoner’s
2.1.1 Participants
Dilemma Game (e.g., Rachlin, 1997; Fantino, Gaitan,
A total of 238 (182 F, 56 M) young adult (M=20.4,
Meyer, & Stolarz-Fantino, 2006) in that both games con-
SD=2.0 years) undergraduate students served as partic-
strain the choices of participants and present them with
ipants. 38 were dropped from the study for either mis-
con?icting options. In that regard the Sharing Game dif-
interpreting the instructions or (in the conditions involv-
fers from the popular UG and DG in that on any given
ing a putative second person) for indicating in debrief-
choice the participant must choose between getting more
ing that during the session, they believed with certainty
for oneself (and still more for the other participant) or
that the second person did not exist.2 Statistical analy-
less for oneself (and still less for the other participant).
ses were carried out with the remaining 200 participants,
Choices on the UG and particularly the DG are relatively
as well as with all 238. The same conclusions and sta-
unconstrained. The Sharing Game forces the participant
tistical ?ndings occurred whether or not these students’
to choose between an outcome that is optimal for both
data were included. Half of the remaining participants re-
participants and one that is competitive (giving a relative,
ceived course credit for volunteering their time; the other
but non-optimal advantage to the allocator). An equi-
half received monetary compensation. Students learned
table choice is absent. However, the trials are arranged
which they were to receive just prior to their sessions.
so that it is possible for the participants to respond eq-
All participants reported being ?uent English speakers,
uitably over trials. Thus, as mentioned above, partici-
free of neurological or psychiatric disorders, and having
pants behaving according to inequality aversion theories
normal or corrected-to-normal vision.
(e.g., Fehr & Schmidt, 1999; Bolton & Ockenfels, 2000)
may be expected to alternate choices, thereby maintain-
ing equal payoffs for both participants. The distribution
2.1.2 Design and stimuli
of choices should permit the characterization of partici-
The economic game developed for this study employed
pants’ choices as optimal, equitable, or competitive. The
a single-player, multiple-trial, two-alternative forced-
possible effects of other variables (such as monetary in-
choice paradigm in which the player’s allocation decision
centive and whether the other participant is a person or
determines both that player’s payoff and that of another
computer) should enable us to determine the extent to
which these distributions of choices are stable or are in-
2Participants’ data were retained 1) if they stated that they were not
?uenced by these variables.
completely certain if the other person existed, but were willing to give
the experimenter the bene?t of the doubt, or 2) if they stated that they
Would participants allocate resources optimally in this
only became certain that there was no other person after completing the
task by always selecting the larger payoff for themselves
allocation session.

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
207
(unseen, passive, and, in fact non-existent) participant.
allocator and recipient in the real money conditions in
Each trial offered participants an opportunity to choose
Experiment 1, respectively, were $8.67 and $9.22.
between two options. One option gave the participants a
Half of the participants were told that the second player
smaller point reward and gave the other player even less.
was an anonymous person in an adjoining room, while
The second option gave participants a larger point reward
the rest were informed that the second player was repre-
and gave the other player even more. To illustrate, a typ-
sented by the computer running the game program. Ta-
ical choice might be:
ble 2 displays the random assignment of men and women
Player One receives 7 and Player Two receives 9
across the conditions of incentive and Player 2 type. Fi-
or
nally, during the game for half the participants, the com-
puter displayed a running tally of both players’ point to-
Player One receives 5 and Player Two receives 3
tals. For the other half, the computer displayed the run-
Table 1 lists the ?ve sets of choices presented to partic-
ning tally only for the ?rst player. The dependent variable
ipants. In each trial, the participant was presented with a
was the percentage of trials in which the participant chose
choice between one of the alternatives in the table’s left
the optimal option, which afforded the maximum amount
hand column and its corresponding alternative in the right
of points for the participant (and, incidentally, for the sec-
hand column. The two options for each choice were al-
ond player).
ways numerically symmetrical in that the absolute value
of the difference between the outcomes for Player 1 and
Table 2: Distribution of men and women across incentive
Player 2 was the same for both alternatives. Regardless
and Player 2 type in Experiment 1
of the particular point amounts offered, participants al-
ways had a choice between the optimal alternative (e.g.,
Men (n = 50) Women (n = 150)
“Player One receives 7 and Player Two receives 9”) and
the competitive alternative (e.g., “Player One receives 5
Monetary
26
74
and Player Two receives 3”). Over 20 trials, the choices
Non-monetary
24
76
were always presented in pairs (e.g., the 7 and 9 ver-
Human
25
75
sus 5 and 3 alternatives were presented twice in a row)
to afford participants a third option: to readily match
Computer
25
75
their earnings with those of the second player. By al-
Monetary/Human
11
39
ternating between the top (optimal) and bottom (compet-
Monetary/Computer
15
35
itive) alternatives, both players would complete the game
with equal (though non-maximal) earnings. For exam-
Non-monetary/Human
14
36
ple, when given the 7 and 9 versus 5 and 3 alternatives
Non-monetary/Computer
10
40
twice in a row, the allocator could a) choose 7 and 9 both
times, resulting in totals of 14 for him- or herself and 18
for the other; b) choose 5 and 3 both times, resulting in
2.2 Procedure
respective totals of 10 and 6; or c) choose 7 and 9 once
and 5 and 3 once, resulting in totals of 12 for each. Each
Participants were assessed individually in a room with
choice pair was presented in random order. Once all ?ve
normal lighting, were seated 50 cm from a personal com-
had been presented (comprising the ?rst ten trials), the
puter running the game program, and were informed that
choice pairs were re-randomized and presented again to
they were to take part in an economic game involving
comprise the remaining ten trials.
resource allocation. The experimenter told participants
The participants who received monetary compensation
that the computer would display multiple trials of differ-
were paid $0.07 USD for every point attained, and were
ent point amounts that the participants could allocate to
told that Player Two would be paid in the same fash-
themselves and to Player 2 (P2), but did not reveal ex-
ion. Participants were also informed that it was possible
actly how many trials there would be or that the choices
to earn up to approximately $10 for themselves. (They
would be presented in pairs. The experimenter verbally
were not told how much the other participant could po-
described how a typical trial would appear, but neither
tentially earn). Those who chose competitively through
suggested any strategy nor explained how the top and
all 20 trials earned $6.72 for themselves and $3.36 for
bottom options were considered optimal and competitive,
the other. Participants who chose optimally every time
respectively. It was at this time that the experimenter ex-
earned $10.08 for themselves and $13.44 for the other.
plained to half of the participants that P2 was an anony-
Those who equalized both participants’ earnings (particu-
mous person in the adjoining room waiting for the ses-
larly those who alternated their choices from trial to trial)
sion to begin (the other half were informed that P2 was
earned $8.40 for both parties. Average earnings for the
the computer). The participant was told that, as Player 1

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
208
Table 1: Choices presented to allocators (Player 1) in the Sharing Game.
First (optimal) alternative
Second (competitive) alternative
Player 1 (participant) Player 2 (putative other or
Player 1 (participant) Player 2 (putative other or
receives
computer) receives
receives
computer) receives
6
8
4
2
6
7
5
4
7
9
5
3
8
11
5
2
9
13
5
1
(P1), only he or she had the ability to choose how much
but now the modes are at 50% (equalizing payoffs) and
both players received in each trial. Participants who were
at 0% (extreme competition). The mode at 100%, seen in
to receive monetary compensation for their involvement
Figures 1a and 1b representing optimal choice, is absent.
in this study were told at this time that each player would
A corollary of these results is that participants chose
receive $0.07 USD for each point they had individually
signi?cantly more optimally in the monetary condition.
accumulated, and that it was possible for P1 to earn up to
This result is shown in Figure 2a, for all participants,
approximately $10 USD (the experimenter did not reveal
and in Figure 2b separately for male and female partic-
how much P2 could potentially earn). Once the exper-
ipants. A three-factor (2 Compensation by 2 Identity of
imenter had determined that the participant understood
P2 by 2 Display salience) analysis of variance was signif-
the verbal instructions, P1 was prompted to read the in-
icant, F(7, 192) = 4.4, p=.0002. The points+money par-
structions displayed on the computer screen. A transcript
ticipants’ tendency to choose optimally was signi?cantly
of the computer-provided instructions is included in the
higher than that of the points-only participants, F(1, 192)
appendix. If participants were in a condition in which
= 25.1, p<.0001. Participants whose points were backed
they were told that P2 was human, the experimenter left
by money chose optimally 59% of the time, while their
the room for approximately 10 s and returned under the
counterparts only did so in 39% of their trials.
pretense of ascertaining that Player 2 was ready. After
making certain that the participant had no questions and
It mattered not at all whether the other participant’s cu-
was ready to begin, the experimenter left the participant
mulative points were displayed for the participant (48%
alone to begin the session.
optimal choices) or not (49%), F(1, 192) = 0.1, ns. When
the analysis revealed that salience had no bearing on per-
2.3 Results and Discussion
centage of optimal choices made, but visual inspection
showed that gender may in?uence behavior, a different
Participants’ choices produced a trimodal distribution
three-factor (2 Compensation by 2 Identity of P2 by 2
with the three modes corresponding to the three straight-
Gender) analysis of variance was conducted. This second
forward strategies: equalizing payoffs, the primary mode,
analysis was signi?cant overall, F(7, 192) = 5.4, p<.0001.
at 50% on Figure 1a; always selecting the optimal op-
Main effects of gender (F(1, 192) = 5.3, p<.03) and in-
tion by choosing the larger payoffs, the second mode at
centive (F(1, 192) = 29.7, p<.0001), as well as a gender-
100% on Figure 1a; always selecting the competitive op-
by-incentive interaction (F(1, 192) = 4.2, p<.05) were re-
tion by choosing the smaller payoffs, the third mode at
vealed. No other signi?cant effects were found. Over-
0% on Figure 1a. When the participants are divided ac-
all, men (n=50) chose optimally 58% of the time while
cording to whether or not points represented monetary
women (n=150) did so 46% of the time. Males were
reward a different picture emerges. Figure 1b, for par-
more in?uenced by the nature of the incentive. As can be
ticipants with monetary incentive, shows a bimodal dis-
seen in Figure 2c, both genders behaved similarly under
tribution with optimal responding now the primary mode,
the points-only condition (38% and 39% for females and
though more participants overall continue to approximate
males, respectively) and both increased their optimality
equalizing payoffs than maximizing payoffs. The third
in the points+money condition, but males (74%) chose
mode from Figure 1a, at 0% representing extreme com-
optimally signi?cantly more often than females (54%).
petition, is absent. Figure 1c shows that when only partic-
Surprisingly, the nature of the other participant did not
ipants without monetary incentive are considered a third
matter in the slightest (F(1, 192) = 0.8, ns): when par-
pattern emerges. Again we have a bimodal distribution
ticipants were told the other participant was a person

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
209
35
35
(a)
fettered determination of the resource allocation in this
study, a result consistent with prior research (van Lange,
30
1999). However in this game that choice also optimizes
25
24
22
the payoff for the second participant (the putative hu-
20
man or computer in fact receives more than the partic-
15
ipant does). Even if our criterion for optimality gener-
Number of subjects
10
ously encompasses choosing the optimal payoff on 75%
5
or more of the trials, only 41 of 200 participants acted
0
optimally. This tendency held even when resources were
0
10
20
30
40
50
60
70
80
90
100
being shared with an inanimate other. Even when mon-
Percent of trials
etary incentive was provided, the other participant was
35
(b)
the computer, and the computer’s cumulative score was
30
not displayed for the participant (the condition in which
25
selecting non-optimally made least sense of all), non-
20
18
optimal choices were made on 41% of the trials.
15
14
Our second central ?nding is that a plurality of our
Number of subjects
10
participants tended to equalize the points allocated to the
5
participant and to the other person or computer. Strict al-
0
ternation between the payoffs occurred for 35 of the par-
0
10
20
30
40
50
60
70
80
90
100
ticipants and 84 of the participants chose optimally on
Percent of trials
between 40% and 60% of trials inclusively. While this
35
(c)
selection pattern is hardly optimal, it is consistent with
30
notions that we have learned rules of fairness (Zizzo &
25
21
Oswald, 2001).
20
19
The instructions for this experiment clearly had a
15
game-playing ?avor that may well have contributed to
Number of subjects
10
the degree of competitive, non-optimal choices made. In-
5
deed, during debrie?ng, 46 participants spontaneously
0
reported that hearing and/or reading the word “game”
0
10
20
30
40
50
60
70
80
90
100
and/or “player” in?uenced their decision to choose com-
Percent of trials
petitively. Thus, Experiment 2 repeated the experiment
with more neutral instructions. The replication would
Figure 1: Number of participants and the percentage of
also afford the opportunity to ascertain whether mon-
the trials in which they chose the optimal option in Exper-
etary rewards would again lead to more optimal deci-
iment 1. 0% indicates never choosing the optimal choice
sions and whether participants’ decisions would again be
(i.e., acting purely competitively); 100% indicates per-
unaffected by the nature of the other player (person or
fect optimal behavior. 50% indicates choosing optimally
computer). Since the running tally of the other partici-
in half of all trials, resulting in non-maximal, but equal
pant’s score (the “salience” variable) proved to have no
(or nearly equal) scores for both players. (a) Data for all
effect in Experiment 1, and since the signi?cant effect
200 participants. (b) Data for participants whose points
of gender was of greater interest to us, we replaced the
were backed by money (n = 100). (c) Data for partici-
salience variable with gender in Experiment 2, equaliz-
pants whose points were not backed by money (n = 100).
ing the number of males and females in each condition.
they chose optimally on 49% of trials; when they were
told the other participant was the computer they chose
3 Experiment 2
optimally on 48% of the trials. This unexpected simi-
larity held across both incentive conditions — those in
3.1 Method
the monetary/human and monetary/computer conditions
3.1.1 Participants
each chose optimally 64% of the time, while those in the
non-monetary/human and non-monetary/computer con-
A total of 156 (84 F, 72 M) young adult (M=20.5, SD=2.1
ditions chose optimally 41% and 37% of the time, respec-
years) undergraduate students served as participants. Six-
tively.
teen women and four men were dropped from the study
The ?rst central ?nding is that participants’ choices
for either misinterpreting the instructions or for being un-
were not consistently optimal even though they have un-
convinced that another person was involved in the exper-

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
210
ically, 17 participants of each gender were randomly as-
100
(a)
signed to the following conditions: a) monetary/human;
b) monetary/computer; c) non-monetary/human; d) non-
80
monetary/computer. As in Experiment 1, statistical anal-
yses were carried out with the remaining 136 participants,
58.8% ***
60
as well as with all 156. The same conclusions and statis-
tical ?ndings occurred whether or not these students’ data
38.6%
40
were included. All participants received course credit for
Percent optimal
volunteering their time; half received monetary compen-
sation for the points they had earned in the course of their
20
session. Students learned which condition (point backed
by money or points alone) they were in just prior to their
0
sessions. All participants reported being ?uent English
Non?monetary
Monetary
speakers, free of neurological/psychiatric disorders, and
having normal or corrected-to-normal vision.
100
(b)
3.1.2 Design, stimuli, and procedure
80
The design, stimuli, and procedure for Experiment 2 were
virtually identical to those used in Experiment 1, with
57.6% *
60
the following exception: all terminology that could be
47.5%
considered as provoking competitive behavior on the part
of the participant was eliminated. When presenting ver-
40
Percent optimal
bal instructions to the participants, the experimenter re-
ferred to the economic game only as a “scenario” or “ac-
20
tivity,” and to the other player as the “other participant,”
the “other person,” or “the computer” (where appropri-
0
ate). In addition, all computer-displayed information was
Women (n=150)
Men (n=50)
re-written to re?ect the same non-provocative language:
“game” was replaced by “scenario,” and “player” by
“person” or “participant.” A transcript of the computer-
100
(c)
provided instructions is included in the Appendix.
Women (n=150)
73.5%
80
Men (n=50)
4 Results and Discussion
53.6%
60
Recall that, in Experiment 1, participants’ choices pro-
38.0%
39.1%
duced a trimodal distribution with the three modes corre-
40
Percent optimal
sponding to the three straightforward strategies: equal-
izing payoffs; always selecting the optimal option by
20
choosing the larger payoffs; always selecting the com-
petitive option by choosing the smaller payoffs (Figure
0
1a). In Experiment 2, however, the trimodal distribution
was replaced by a bimodal one as shown in Figure 3a.
Non?monetary
Monetary
By eliminating the instructions suggesting a game, we
Figure 2: Percentage of trials in which participants chose
also eliminated the mode above 0% indicating extreme
the optimal over the competitive option as a function of a)
competitiveness. As in Experiment 1 the primary mode
reinforcement, b) gender, and c) reinforcement by gender
was at 50%, suggesting a tendency to equalize payoffs.
in Experiment 1. (Table 2 lists the distribution of men
A secondary mode is seen at 100%, indicating choice of
and women across the conditions displayed in 2c.) Error
the optimal option. As in Experiment 1, the nature of
bars are ± SEM. * p < .05. *** p < .0001.
the incentive made an important difference. As shown
in Figure 3b, when points were exchangeable for money,
a unimodal negatively skewed distribution emerged, with
iment (as determined during debrie?ng), leaving 68 of
the mode at 100% (i.e., choosing optimally in every trial).
each gender evenly distributed in all conditions. Specif-
When earned points were not exchangeable for money

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
211
35
(a)
mally more frequently (64% of trials) than women (54%);
32
this result is displayed in Figure 4b. However, the signi?-
30
cant incentive/gender interaction found in the prior exper-
25
iment did not replicate: men were not signi?cantly more
20
16
in?uenced by the nature of the incentive, F(1, 128) = 0.2,
15
ns. As can be seen in Figure 4c, males’ and females’
Number of subjects
10
means in the points condition were 54% and 46%, respec-
5
tively, while their means in the points+money condition
0
were 74% and 62%, respectively.
0
10
20
30
40
50
60
70
80
90
100
As in Experiment 1 there was no difference in the re-
Percent of trials
sults as a function of the nature of the other participant
35
(b)
(person or computer). When participants were told that
30
the other participant was a person they chose optimally
25
on 61% of trials; when they were told that the other par-
20
ticipant was the computer they chose optimally on 57%
15
13
of the trials. This difference was not signi?cant (F(1,
Number of subjects
10
128) = 1.1, ns) and was in the opposite direction of what
5
we had expected (more optimal decisions when the other
0
participant was a computer). As with the previous experi-
0
10
20
30
40
50
60
70
80
90
100
ment, this similarity held across both incentive conditions
Percent of trials
— those in the monetary/human and monetary / com-
35
(c)
puter conditions, respectively, chose optimally 70% and
30
26
66% of the time, while those in the non-monetary/human
25
and non-monetary / computer conditions chose optimally
20
52% and 47% of the time, respectively.
15
Thus, the results of Experiment 2 replicated those of
Number of subjects
10
Experiment 1 in all important respects, including the sig-
5
ni?cant effects of monetary incentive and of gender, the
0
tendency of a plurality of students to equalize payoffs,
0
10
20
30
40
50
60
70
80
90
100
and the lack of an effect based on the nature of the other
Percent of trials
participant (person or computer). However, the nature
Figure 3: Number of participants and the percentage of
of the instructions differed for the two experiments and
the trials in which they chose the optimal option in Exper-
this difference markedly affected the results. Speci?-
iment 2. 0% indicates never choosing the optimal choice
cally, when aspects of the instructions suggesting a game
(i.e., acting purely competitively); 100% indicates per-
were removed, the likelihood of competitive choices was
fect optimal behavior. 50% indicates choosing optimally
altered dramatically (compare Figures 1a and 3a, left
in half of all trials, resulting in non-maximal, but equal
sides).
(or nearly equal) scores for both players. (a) Data for all
136 participants. (b) Data for participants whose points
were backed by money (n = 68). (c) Data for participants
5 General discussion
whose points were not backed by money (n = 68).
The sharing task studied in the present experiments of-
fered participants repeated binary choices in which the
there was a unimodal distribution, with the mode at 50%
payoffs for one outcome pair were higher for both play-
(Figure 3c). A three-factor (2 Compensation by 2 Identity
ers (and the chooser received the smaller payoff) and the
of P2 by 2 Gender) analysis of variance was signi?cant,
payoffs for the other outcome were lower for both play-
F(7,128) = 3.3, p<.003. Main effects of incentive (F(1,
ers (and the chooser received the larger payoff). On any
128) = 16.6, p<.0001) and gender (F(1, 128) = 5.0, p<.03)
given trial the participant was constrained to select be-
were revealed, but there were no signi?cant interactions.
tween these two outcomes. Unlike the well-established
No other signi?cant effects were found. In a pattern very
Ultimatum and Dictator games, which do not constrain
similar to that seen in Experiment 1, participants whose
participants’ choices in this way, the Sharing Game al-
points were backed by money chose optimally 68% of
lows us to delineate between those who prefer to max-
the time, while their counterparts only did so in 50% of
imize their earnings and those who prefer a maximized
their trials (see Figure 4a). Once again, men chose opti-
relative advantage over the other.

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
212
a computer did not matter. Using such terminology as
100
(a)
“game” and “player” in the in the ?rst experiment was
associated with an increased frequency of competitive be-
80
havior relative to the second experiment, which employed
67.8% ***
more neutral language. Males chose optimally more
60
frequently than females (a gender-by-incentive interac-
49.9%
tion was revealed in the ?rst experiment), and in agree-
40
ment with prior research, participants in the monetary
Percent optimal
conditions acted more optimally than those in the non-
monetary conditions. Overall, equality was the modal
20
strategy employed.
Participants in our experiments did not show a differ-
0
ence in the way they treated a P2 they believed to be hu-
Non?monetary
Monetary
man versus a P2 they knew to be represented by a com-
puter. The Computers Are Social Actors paradigm (al-
100
ternately known as Social Responses to Computer Tech-
(b)
nologies, or SRCT) posits that people respond to com-
puters in a social manner nearly equal to how they re-
80
spond to other humans, and our results appear to support
63.8% *
this. However, previous studies involved having the com-
60
53.9%
puter as more of an active partner rather than a silent ob-
server serving mostly as a data input port. The computer
40
partners used to study the CASA model were also pro-
Percent optimal
grammed to display some aspect of human personality, be
it a male or female voice (Nass, Moon, & Green, 1997),
20
an animated face of a certain ethnicity (Nass, Isbister,
& Lee, 2000), or emitting humorous remarks during the
0
course of a task (Morkes, Kernal, & Nass, 1999). Unlike
Women (n=68)
Men (n=68)
the computers in the aforementioned studies, our comput-
ers did not truly interact with their users, nor were they
100
programmed to simulate any personality characteristics:
(c)
Women (n=68)
they did nothing beyond displaying textual instructions,
Men (n=68)
73.7%
80
visual information of what choices were available, how
61.9%
much was earned, and when the session was ?nished.
53.8%
Although our results do not answer the question of why
60
45.9%
people respond to a computer the same way as to another
person, they at least suggest that a computer need not im-
40
itate an overt aspect of human personality to elicit similar
Percent optimal
responses to it (probably the most personable thing our
computers did was to preface its text instructions with the
20
word, “Welcome!”). Despite these differences, the results
are consistent with prior studies in showing no difference
0
between how a participant acts upon the other participant
Non?monetary
Monetary
(human or computer).3
Note also that deception cannot be an explanation of
Figure 4: Percentage of trials in which participants chose
the optimal over the competitive option as a function of a)
3Of those in the monetary/computer conditions, 18 participants from
reinforcement, b) gender, and c) reinforcement by gender
Experiment 1 and 10 from Experiment 2 ended up with a greater number
of points than the computer. In Experiment 1, 15 of those 18 explicitly
in Experiment 2 (there are 34 participants in each of the
stated that they wanted more points than the computer, but only 5 of
four conditions displayed in 4c). Error bars are ±1 SEM.
them indicated that the “game” and “player” language in?uenced them
* p < .05. *** p < .0001.
to behave that way. In Experiment 2, 6 of the 10 directly indicated that
they wanted to have a relative advantage over the computer. For exam-
ple, one participant said, “I didn’t want the computer getting more than
me.” Another, in the monetary/human condition said that if she were
Across both experiments, whether participants were
paired with a computer instead, she “would not care about its feelings;
told that the anonymous other was another student or
I would just want to beat it.”

---

Judgment and Decision Making, Vol. 2, No. 3, June 2007
Fairness in resource allocation
213
the result that some subjects sacri?ced their own gains in
It is conceivable that the gender/incentive interaction
order to achieve a relative advantage over the computer,
observed in Experiment 1 may be part of a larger, three-
since no deception was involved for participants in the
way interaction involving “game” terminology; i.e., be-
computer conditions.
ing primed by the words “game” and/or “player” might
The lack of difference between the computer and hu-
in?uence men in the monetary conditions to act more op-
man conditions could arguably be due to participants not
timally, or in?uence women in the same conditions to act
believing that the other was really there, but among those
less so. While there was no signi?cant interaction in the
in the human conditions, 72 of the 100 in Experiment 1
neutral-language Experiment 2, the data were trending
and 32 of the 68 in Experiment 2 did not indicate dur-
in the same direction as in the ?rst experiment. Since
ing debrie?ng that they even suspected that the other per-
a direct statistical comparison between the different uses
son was not real. The remaining 28 from the ?rst exper-
of language is not feasible in the current experiments, it
iment and 36 from the second stated that although they
could be quite informative to conduct a future Sharing
weren’t absolutely certain that they were paired with an-
Game study incorporating gender, incentive type, and ter-
other, they allowed for the possibility of such, and made
minology as the independent variables.
their choices accordingly (participants who stated that
Although it could be argued that presenting the same
they were certain that no other person was involved were
choice twice in a row may have induced people to make
dropped). Additionally, of those in the human recipient
one choice ?rst and the other one second, we suggested
conditions, 21 from Experiment 1 and 38 from Experi-
no such strategy (nor any other strategy) to our partic-
ment 2 stated that had their human recipients been com-
ipants, nor did we tell them that each choice would be
puters instead, they would have made different allocation
repeated. Also, while 44 participants in Experiment 1
choices, suggesting that they did perceive the other as a
and 39 participants in Experiment 2 expressed a desire
real person. Excluding those who showed any doubt as to
for equality, only 22 and 15, respectively, alternated their
the other person’s existence from our analyses did not af-
choices to achieve this goal. Additionally, of those who
fect the (non-)signi?cance of the human/computer result
alternated, none of them reported doing so simply be-
in either experiment.
cause of the pattern being presented to them. They may
In their meta-analysis of gender and competition, Wal-
have stated that they saw the pattern, but they also stated
ters, Stuhlmacher, and Meyer (1998) stated that the con-
that they were seeking equality anyway. Zero participants
tention that women behave more cooperatively than men
said that the alternation pattern they had discovered in?u-
had support, albeit limited: overall, “gender accounted
enced them to choose equitably.
for less than 1% of the variance (r2) in negotiator com-
Bazerman et al. (1992) reported data from somewhat
petitiveness.” The authors went on to state that in studies
related allocation decisions, albeit ones in which an eq-
wherein partners had little contact with each other and
uitable option was included. In a typical item partici-
the negotiations did not proceed beyond making choices
pants were asked to choose between $500 for themselves
to cooperate or compete, gender differences were virtu-
and another person or between $600 for themselves and
ally zero. This stands in contrast to the results of our
$800 for the other person. In several conditions in each
studies, in which men consistently behaved more opti-
of two studies they found strong preferences for the more
mally (in an strictly economic sense) than women, who
optimal outcome (here, $600 / $800). There was little
were more prone to sharing equally. However, their meta-
evidence in their studies of the preference for equitable
analysis concentrated on matrix games such as the Pris-
outcomes found in the present studies. There are several
oner’s Dilemma, in which each player makes moves that
differences between their experiment and ours including
directly in?uence both people’s welfare. While our ex-
the following: their participants were business students
periments were similar to those analyzed by Walters and
and accounting ?rm managers, whereas ours were under-
colleagues with respect to the lack of contact between
graduates; they provided scenarios with their questions
players (our second person was nonexistent), our partici-
whereas we did not; they did not repeat the same trials
pants made all the decisions that in?uenced both partners’
whereas we did; the reward amounts they employed were
welfare. It should be noted that Walters et al. (1998) em-
approximately one hundred-fold in magnitude greater
phasized their ?nding that the gender differences they re-
than ours; they used only hypothetical rewards whereas
vealed were, although signi?cant, quite small, and prone
we used both hypothetical and actual monetary rewards.
to attenuation or even reversal, depending on the condi-
The results of the present experiment point to the role of
tions of the study. Our own results are in keeping with
procedural and contextual factors in the allocation deci-
this conservative outlook; although we found consistent
sions of our participants (for example, the nature of the
gender differences in our studies, we also revealed that
incentives and the instructions). In that light and in light
context can in?uence one to act more or less competi-
of the many differences in the procedures of the studies
tively in an allocation scenario.
of Bazerman et al. (1992) and of the present studies, it

---