Prosocial Lie Selfish Honesty

When deception promotes trust
Maurice E. Schweitzer
1
2
Emma Levine
3
To me it seems certain that every lie is a sin…
–St. Augustine
Deception deprives people of two of the most
principal instruments for interpersonal action—
trust and belief.
–Sir Francis Bacon
By a lie, a man annihilates his dignity
–Immanuel Kant
Deception is unethical
–Chuck Klosterman
NYTimes, The Ethicist, Jan 2014
4
Virtue of Honesty:
Embedded in Modern Organizations
“We are honest.”

Dell Inc., Code of Conduct, 2015
“Honesty, Fairness, Integrity and Convenient Foods.”

5
Pepsi Co., Mission Statement
Deception and trust
“Deception causes enduring harm to trust.”

Schweitzer, Hershey, & Bradlow, 2006

Santoro & Paine, 1993; Carr, 1968; O’Connor & Carnavale, 1997;
Schweitzer & Croson, 1999; Steinel & De Dreu, 2004; Larzelere &
Hutson, 1980; Lieberman, 1981; Ring & Van de Ven, 1992
Honesty is a critical antecedent of trust


6
Aquino &Reed, 2002; Lapsley & Lasky, 2001; Walker & Hennig,
2004; Davis, Mayer & Schoorman 1995; Butler & Cantrall, 1984
Selfish
Deception
7
Prosocial
Honesty
1) How do prosocial lies influence
trust?
2) How does deception, per se,
influence trust?
8
Prosocial lies

Statements told with the intention of misleading
and benefitting other(s)

Prosocial lies are:



Learned at a young age
Often told as a form of politeness
Told in roughly 20% of social interactions
DePaulo & Kashy, 1996
Prosocial lying and trust
10
Rousseau, Sitkin, Burt, & Camerer, 1998; Mayer, Davis, Schoorman, 1990
Prosocial lying and trust
11
Rousseau, Sitkin, Burt, & Camerer, 1998; Mayer, Davis, Schoorman, 1990
Prosocial lying and trust
12
Rousseau, Sitkin, Burt, & Camerer, 1998; Mayer, Davis, Schoorman, 1990
Today’s Talk

Study 1: Experienced prosocial lies and
trust

Study 2: Observed prosocial lies and
trust

Study 3: Disentangling deception from
intentions
13
Study 1:
Experienced prosocial lies

Mturk (N=121)
Stage 1: Play a game with a confederate
Stage 2: Learn that the confederate was
either selfishly honest or told a prosocial lie
Stage 3: Make trust decisions
14
Stage 1: Play a modified deception game
Gneezy, 2005

Participants are RECEIVERS, paired with
Confederate SENDERs.
A coin is flipped
The Confederate (SENDER)
has to report the outcome
to the Participant
(RECEIVER)
The Participant
(RECEIVER) reports
the outcome to the
experimenter
“The coin landed on HEADS.” (Honest)
“The coin landed on TAILS.” (Lie)
15
The Confederate
(SENDER) and
Participant (RECEIVER)
are paid
Stage 2:2:
Play
a deception game
Stage
Participants
learn that the confederate
either was honest or lied

Participants learn that the confederate had the following
choices:
Selfish Honesty
Prosocial Lie
Send the message:
“The coin landed on HEADS.”
Send the message:
“The coin landed on TAILS.”
Payoffs are:
• $2 to the SENDER (Confederate)
• $0 to the RECEIVER (Participant)
Payoffs are:
• $1.75 to the SENDER (Confederate)
• $1 to the RECEIVER (Participant)
16
Stage 3: Modified Trust Game
Berk, Dikhaut, & McCabe, 1995
Confederate
Participant
17
Stage 4: Attitudinal measures

Trusting attitudes (r = .89)



Perceived Benevolence (α=.92):


I trust my partner to RETURN money
I am confident that my partner will RETURN money
How kind, nice, selfish (reverse coded) is your partner?
Perceived Deception (α=.87):



18
My partner sent an honest message in EXERCISE 1 (reverse
coded).
My partner lied about the outcome of the coin flip in
EXERCISE 1.
My partner deceived me in EXERCISE 1.
Results:
19
(% of participants who passed in the trust game)
Trusting Behavior
Results: Passing in the Trust Game
56%
32%
Honesty
20
Prosocial Lie
χ2 (1, N = 121) = 6.88, p < .01
Results: Attitudinal Trust
6
4
(1-7)
Attitudinal Trust
5
3
2
1
Honesty
21
Prosocial Lie
F(1, 119) = 9.85, p < .01; d = .55
Results: Perceived Benevolence
Perceived Benevolence
(1-7)
6
5
4
3
2
1
Honesty
22
Prosocial Lie
F(1, 119) = 8.12, p < .01; d = .50
Results: Perceived Deception
Perceived Deception
(1-7)
6
5
4
3
2
1
Honesty
23
Prosocial Lie
F(1, 119) = 102.60, p < .001; d = 1.35.
What about different types of trust?
24
Study 2:
Observed prosocial lies

Mturk (N=139)

Same design as study 1, except:


25
participants simply observe the deception game, and
deception is not costly for the liar
Trusting Behavior
(% of participants who passed in the trust game)
Study 2: Passing in the Trust Game
69%
37%
Honesty
26
Prosocial Lie
χ2 (1, N = 139) = 13.65, p < .01
What about different types of trust?
How do we judge deception, per se?
27
Study 3:
Disentangling deception and intentions

Wharton Behavioral Lab (N=312)

Same design as study 2, except
2(Decision: honesty vs. lie) x
2(Intentions: prosocial vs. selfish)
28
Study 3:
Disentangling deception and intentions
Participants observe a confederate who makes one of the
following 4 choices:
Be honest…
•
•
Selfish Honesty
$2 to the SENDER (Confederate),
$0 to the RECEIVER
29
Lie….
•
•
Prosocial Lie
$1.75 to the SENDER (Confederate),
$1 to the RECEIVER
Study 3:
Disentangling deception and intentions
Participants observe a confederate who makes one of the
following 4 choices:
Be honest…
•
•
$2 to the SENDER (Confederate),
$0 to the RECEIVER
Be honest…
•
•
Selfish Honesty
Prosocial Honesty
$2 to the SENDER (Confederate),
$0 to the RECEIVER
30
Lie….
•
•
$1.75 to the SENDER (Confederate),
$1 to the RECEIVER
Lie….
•
•
Prosocial Lie
Selfish Lie
$1.75 to the SENDER (Confederate),
$1 to the RECEIVER
Trusting Behavior
(% of participants who passed in the trust game)
Study 3: Passing in the trust game
Honesty
47%
26%
48%
24%
Selfish Intentions
31
Lie
Prosocial Intentions
Effect of Intentions: χ2 (1, N = 312) = 16.70, p < .01. Effect of outcome: ns.
Study 3: Perceived Deception
Perceived Deception
(1-7)
7
Lie
6
5
4
3
2
1
32
Honesty
Selfish outcome
Selfish
Intentions
Prosocial Outcome
Prosocial Intentions
Note. Main effect of lying: p < .001. Main effect of outcome: ns.
Conclusions and Implications

We challenge the assumption that deception harms
trust

Benevolence is often more important than honesty

It is important to disentangle deception from selfserving outcomes and intentions
33
34
Related findings

What about a control?



What about different types of trust?


Lies told to protect others’ feelings (rather than generate
monetary gains) have the same effects
What about repeated behavior?

35
Prosocial lying increases emotional attachment (affective
trust), but decrease integrity-based trust
What about different types of lies?


Prosocial (lying) increases trust relative to control
conditions;
Selfish (honesty) decreases trust relative to control
conditions
Always being kind is better than always being honest
(% of participants who passed in the trust game)
Trusting Behavior
What about a control condition?
71%
44%
28%
Selfish
Honesty
22%
Selfish
Lie
Selfish Decision
36
66%
49%
Control
1
Control
2
Alt. lie
Selfish lie
Control (No Information)
Altruistic Altruistic
Lie
Honesty
Altruistic Decision
What about reciprocity?
Trusting Behavior
(% of participants who passed in the trust game)
Honesty
69%
59%
36%
Mutually Beneficial Lie
37
Lie
χ2 (1, N = 154) = 7.78, p < .01
37%
Prosocial Lie
χ2 (1, N = 139) = 13.65, p < .01
What about different types of trust?
7
Honesty
6
Prosocial Lie
p < .01
5
Trust
(1-7)
p < .01
4
3
2
1
Affective Trust
e.g. “I would rely on this
person for support when I
need it.”
Integrity-based Trust
e.g. “I would expect this
person to give me accurate
feedback on an idea.”
Scenario Study, N = 303,
Stimuli38are prosocial lies participants told (e.g. “the food was great” when it was overcooked)
What about repeated behavior?
Attitudinal Trust
(1-7)
6
5
4
3
2
1
Always Selfish Always dishonest Always Honest Always Prosocial
IV: observe 4 rounds of the deception game, payoff structure switches
39
N = 150, all cells are different at p < .02
What about a control condition?
Payment Structure:
(% of participants who passed in the trust game)
Trusting Behavior
Altruistic Lie-Selfish Honesty
71%
44%
28%
Selfish
Honesty
66%
49%
22%
Selfish
Lie
Selfish Decision
40
Altruistic Honesty-Selfish Lie
Control
1
Control
2
Control (No Information)
Altruistic Altruistic
Lie
Honesty
Altruistic Decision
Prosocial lying is perceived as moral
7
*
Moral character
6
*
5
4
3
2
1
Altruistic Truth
Selfish Lie
Typical contrast:
Altruistic truth vs. Selfish lie
Selfish Truth
Altruistic Lie
Focal contrast:
Selfish truth vs. Altruistic lie
* p < .05.
% Predicting that the target will tell the
truth when the payoffs are switched
Prosocial lying is a better signal of future
honesty
68%
48%
Selfish Honesty
42
Prosocial Lie
N = 80; p = .06
APPENDIX
43
Study 2:
Prosocial and mutually-beneficial lies

Mturk (N=293)

Same design as study 1, except participants
simply observe the deception game and
2(Decision: honesty vs. lie)
x 2(Lie: prosocial vs. mutually beneficial)
44
Study 2:
Prosocial and mutually-beneficial lies
Participants observe a Confederate that makes one of the
following 4 choices:
Be honest…
•
•
$2 to the SENDER (Confederate)
$0 to the RECEIVER
Be honest…
•
•
Antisocial Honesty
Mutually-harmful
Honesty
$2 to the SENDER (Confederate)
$0 to the RECEIVER
45
Lie….
•
•
$2 to the SENDER (Confederate)
$1 to the RECEIVER
Lie….
•
•
Prosocial Lie
Mutually-beneficial Lie
$2.25 to the SENDER (Confederate)
$1 to the RECEIVER
Study 2: Trusting behavior
Trusting Behavior
(% of participants who passed in the trust game)
Honesty
69%
59%
36%
Mutually Beneficial Lie
46
Lie
χ2 (1, N = 154) = 7.78, p < .01
37%
Prosocial Lie
χ2 (1, N = 139) = 13.65, p < .01
Study 2: Perceived Deception
Perceived Deception
(1-7)
6
Honesty
Lie
5
4
3
2
1
Mutually Beneficial Lie
47
Prosocial Lie
Note. Main effect of deception: p < .001. Main effect of type of lie : n.s. Interaction: n.s.
48
Study 3: Perceived benevolence
Perceived Benevolence
(1-7)
6
Honesty
Lie
5
4
3
2
1
Selfish outcome
Prosocial Outcome
49
Note. Main effect of outcome: p < .001. Main effect of lying: p < .001.
Study 3: Perceived Deception
Perceived Deception
(1-7)
7
Lie
6
5
4
3
2
1
50
Honesty
Selfish outcome
Selfish
Intentions
Prosocial Outcome
Prosocial Intentions
Note. Main effect of lying: p < .001. Main effect of outcome: ns.
Study 3: Attitudinal trust
Attitudinal Trust
(1-7)
6
Honesty
Lie
5
4
3
2
1
Selfish outcome
Prosocial Outcome
51
Note. Main effect of outcome: p < .001. Main effect of lying: p =.02
Study 3: Affective vs. Integrity-based Trust

Amazon Mechanical Turk, N=302
Design:
2 (prosocial lie vs honesty) x
3 (different lies as stimuli) X
2 (gender of target)

52
Pilot Study


100 Mturk participants listed 5 lies each
Rated each lie on:


53
I told this lie to benefit myself.
I told this lie to benefit another person or
people.
Pilot to Study 3: Examples

Lies to benefit others:






I told my boyfriend his haircut looked nice, but it looks funny
I told my wife that she wasnt fat.
I told someone that their sweater looked nice.
I told my girlfriend she looked good when she really looked terrible
i told my husband i had a migraine to get out of sex.
Lies to benefit self:





54
I suggested to an employee that she apply for a transfer but I am not likely
to approve it.
I didn't tell my friends the complete truth about why my roommate moved
out.
i made my resume look a lil better
I told my roommate I didn't pour the coffee down the drain but I did.
i told my friend i did not have sex with his wife but i did
Pilot to Study 3: Other results




Lies told to benefit the self are often told to minimize
conflict
Lies told to benefit others are often told to be polite
People only feel bad about telling lies that benefit the
self
Gender effects:

55
Men agree significantly more with the statement: “honesty is
more important than kindness.” (p<.01)
Results: Affective Trust
Affective Trust
(1-7)
6
5
4
Honesty
Prosocial Lie
3
2
1
56
You [don't] look
nice
The food is [not] I [don't] like the
good
gift
Alpha=.90
Effect of lie, collapsed across stimuli; F(1,301) = 52.05, p< .01
Results: Integrity-based trust
Integrity-based Trust
(1-7)
6
5
4
3
2
1
57
Honesty
Prosocial Lie
You [don't] look
nice
Alpha=.85
The food is [not]
good
I [don't] like the
gift
Main effect of lie, collapsed across stimuli; F(1,301) = 284.50; p < .01
Study 4:
Disentangling prosociality from deception

Amazon Mechanical Turk, N=241
Design:
2 (Deception: honesty vs. lie) x
3 (Motive: prosocial vs. no motive) x
2 (Stimulus sampling)

58
Study 4: Stimuli
Scenario 1: Presentation
John watched you give a presentation.
Lie:
The presentation was great!
…even though he noticed several mistakes
Scenario 2: Suit
You met with Jane before an important meeting. You
just bought a new suit.
Lie:
You look great!
…even though she thought you looked unprofessional
boost confidence vs none
Motive:
boost confidence vs none
Honesty:
You made several mistakes.
Honesty:
You look unprofessional.
Motive:
Motive:
Improve in future vs none
Motive:
Improve in future vs none
59
Study 4: Measures
Affective Trust
(Dunn, Ruedy, & Schweitzer, 2012)



I would talk with this person about difficulties I am having at school.
I would be willing to admit my worst mistakes to this person.
I would rely on this person for support when I need it.
Integrity-based Trust
(items from Johnson-George and W. C. Swap, 1982)



Jane would tell me the truth if I asked for feedback on an idea related to my job.
If Jane gave me a compliment on my haircut I would believe Jane meant what was said.
Jane would not lie.
Manipulation check


To what extent was Jane/John honest?
To what extent did Jane/John express his/her true opinion?
60
Study 4: Measures
Mechanisms:
 Integrity (Ferrin, Kim, Cooper, & Dirks, 2007)
 Jane will stick to her word
 Jane has a great deal of integrity
 Jane cares about honesty and truth
 Benevolence
 Jane has good intentions
 Jane is benevolent
 Jane would not purposefully hurt others
61
Results: Affective Trust
6
Presentation
Honesty
Lie
Suit
Affective Trust
(1-7)
5
4
3
2
1
No Motive
Prosocial Motive
No Motive
Prosocial Motive
Note. Main effect of motive: p < .001. Marginal main effect of lying: p = .07.
Results: Integrity-based Trust
6
Presentation
Honesty
Lie
Suit
Integrity-based Trust
(1-7)
5
4
3
2
1
No Motive Prosocial Motive
63
No Motive
Prosocial Motive
Note. Main effect of lying: p < .001. Main effect of motive: ns.
Study 5: Mediation
Perceived
Benevolence
Prosocial
Intentions
Trusting Behavior
Lie Condition: Indirect Effect of benevolence = 1.14, SE = .25; 95% CI [0.70, 1.67]
Honesty condition: Indirect Effect of benevolence = .97, SE = .23; 95% CI [0.58, 1.44]
64
FAQs

Is this an artifact of the experiment?




This is recognized as deception; prior studies show lie aversion for
even these types of lies
Scenario studies also document benefits of lying
Not all lies will have this effect, but there does exist a category of
deceptive behavior that has positive rather than the assumed
negative consequences on character judgments and interpersonal
trust
Only care about the money?


Only 33% tell prosocial lie; same % tell the selfish lie….204 participants
separate pilot
Reciprocity, giving them money they gave up, etc.


65
Trusting attitudes – not just giving them money, they indicate an
expectation to repay
Benefit to others – benevolence