Good Lamps Are the Best Police: Darkness Increases Dishonesty and Self-Interested Behavior (Zhong, Bohns, Gino)

Core ideas and theoretical backdrop

• Darkness can do more than simply make anonymity possible; it can create an illusory sense of anonymity that licenses dishonest and self-interested behavior, even when actual anonymity is preserved.
• The study connects historical and theoretical ideas about darkness and conduct:
  • Emerson (Worship) suggested darkness acts as a social regulator by exposing acts to the world; here darkness is proposed to disinhibit behavior via perceived concealment.
  • Plato’s Ring of Gyges as a philosophical precedent for anonymity leading to wrongdoing;
  • Deindividuation literature (e.g., Festinger, Pepitone, & Newcomb, 1952; Zimbardo, 1969) links anonymity and group dynamics to disinhibited behavior.
• Distinction from classic deindividuation: the authors argue darkness produces illusory anonymity that is anchored in one’s phenomenological experience (subjective perception), not just objective cues of anonymity.
• The central mechanism tested across three experiments is perceived anonymity as a mediator between darkness and unethical/self-interested actions.
• Key concept: illusory anonymity (not actual anonymity) under darkness can license dishonesty and self-serving choices by altering people’s beliefs about others’ ability to observe them and by triggering egocentric biases in perspective-taking.
• Piagetian idea of egocentrism in children is used to illustrate how adults may still assume others share their limited perspective under darkness.
• Real-world relevance: implications for urban lighting, consumer contexts, and everyday ethical decisions when visibility is low.

Definitions and core constructs

• Darkness and dim lighting as a manipulation of environmental illumination.
• Illusory anonymity: a felt sense of being hidden or unobserved that is not grounded in actual concealment of one’s actions.
• Actual anonymity: information about actions that is not linked to identity or that cannot be traced back to the actor.
• Dishonesty and self-interested behavior: actions that prioritize one’s own gain at the expense of truth or fairness (e.g., overreporting performance, making unfair offers).
• Perceived anonymity as a mediator: subjective sense of being unseen predicts behavioral outcomes, potentially explaining why darkness increases unethical behaviors.

Experiment 1: Cheating in a Dim Room

• Objective: test whether slight environmental dimness increases cheating, beyond guaranteed anonymity.
• Design: randomized between-subjects (dim room vs control room); anonymous, individual task with identifiable self-report bias potential.
• Participants: 84 undergraduates at University of North Carolina at Chapel Hill (UNC Chapel Hill): 40 female, 44 male; average age 20.54; compensated up to $12 total ($2 show-up fee + up to $10 additional).
• Task and setup:
  • Two sheets of paper: a worksheet with 20 matrices (each matrix has three-digit numbers) and a collection slip for reporting results; back of the slip included a different matrix as an example.
  • Goal: find two numbers in each matrix that sum to 10 within 5 minutes.
  • Payment: for each correctly identified pair, participants keep $0.50; the rest of the $10 is transferred to a white envelope; the date and nameless collection slip allow matching to compute overreporting.
  • The task is designed to be anonymous: no identifying information on sheets; one unique number on the test sheet matches a number on the collection slip to enable matching but not to identify cheating beyond the study’s scope.
• Conditions: dim room (lights out for some) vs well-lit control room; dim room made by reducing the number of ceiling lights (12 -> 4) but participants could still see each other.
• Main dependent variable (cheating measure):
  • Difference score = self-reported performance − actual performance; positive scores indicate overreporting/cheating.
• Actual performance results:
  • Dim room: M = 7.26, SD = 2.27
  • Control room: M = 6.95, SD = 2.49
  • Statistical test: t(82) < 1, ext{p} = .56, ext{prep} = .46 (no significant difference in actual performance)
• Self-reported performance results:
  • Dim room: M = 11.47, SD = 4.32
  • Control room: M = 7.78, SD = 3.09
  • Statistical test: t(82) = 4.48, p < .001, ext{prep} > .99
• Cheating outcomes:
  • Difference in self-reported vs actual performance (cheating) by condition:
  • Average overstatement: Dim room M = 4.21, SD = 4.12 vs Control room M = 0.83, SD = 1.58; t(82) = 4.92, p < .001, ext{prep} > .99
  • Proportion of participants who overstated: Dim room 60.5% (SD = 50%) vs Control room 24.4% (SD = 44%); ext{ } \chi^2(1, N = 84) = 11.15, p = .001, ext{prep} = .99
• Conclusion from Experiment 1:
  • Slight room dimness increases cheating relative to a well-lit room, even when absolute anonymity is guaranteed.
  • Darkness has a robust effect on dishonesty beyond just creating anonymity or making it easier to cheat under concealment.
• Additional notes and context:
  • The design isolates darkness as the causal factor by ensuring anonymity and controlling time to solve matrices; prior studies varied dimness with other confounds (e.g., confidentiality of personal information) and did not isolate dimness as a causal factor.

Experiment 2: “Shades” and Self-Interested Behavior

• Objective: test whether wearing sunglasses (illusory anonymity) increases selfishness in an anonymous online interaction.
• Design: one-factor between-subjects (sunglasses vs clear glasses).
• Participants: 50 university students at University of Toronto (31 female, 19 male); average age 21.36; maximum payment $11 (5 for showing up; up to $6 in the study).
• Procedure:
  • Participants randomly assigned to wear oversized sunglasses or clear glasses; no face-to-face interaction; interactions are online via a dictator game setup.
  • Task: one-shot dictator game mediated by a computer; initiator (the participant) has $6 to share with a recipient; recipient can accept or reject, but their decision cannot affect the initiator’s payoff—this ensures anonymity of action visibility.
  • The sunglasses condition aims to induce illusory anonymity even though there is no impact on actual visibility in the online setup.
• Dependent variable: amount offered to the recipient (0–$6).
• Results:
  • Sunglasses condition: M = $1.81, SD = $1.30
  • Clear glasses condition: M = $2.71, SD = $1.83
  • Statistical test: t(48) = 2.02, p = .049, ext{prep} = .88 (significant but small effect)
  • Comparison to fair division ($3): sunglasses vs fair division: t(25) = -4.688, p < .01, ext{prep} > .95 (sunglasses offers are significantly less than fair share)
  • Clear glasses vs fair division: t(23) = 0.78, p = .44, ext{prep} = .54 (no significant difference from fair division)
• Conclusion from Experiment 2:
  • Wearing sunglasses leads to more selfish offers in an online, anonymous dictator game, consistent with the illusory anonymity hypothesis.
  • The effect is not due to actual changes in perception by others, given the lack of face-to-face interaction.
• Notes:
  • The experiment was designed to isolate illusory anonymity from actual anonymity, using an entirely computer-mediated task.

Experiment 3: “Shades” and Perceived Anonymity as a Mediator

• Objective: directly test whether perceived anonymity mediates the effect of darkness (sunglasses) on selfish offers observed in Experiment 2.
• Design and procedure: identical to Experiment 2 with the following addition:
  • A five-item perceived anonymity scale (Table 1) administered after the dictator game to capture the extent to which participants felt anonymous and believed others were paying attention to their behavior.
  • Scale: 7-point Likert (1 = strongly disagree, 7 = strongly agree); reverse-scored items noted as *; internal consistency: α = .93.
  • Items assess: being watched, being anonymous, my choice going unnoticed, my identity not known, others paying attention.
• Participants: UNC Chapel Hill undergraduates; N = 83 (39 female, 44 male); average age 20.71; maximum payment $11 ($5 show-up + up to $6).
• Behavioral results (dictator game offers):
  • Sunglasses: M = $1.93, SD = $1.27
  • Clear glasses: M = $2.76, SD = $1.46
  • Statistical test: t(81) = -2.77, p < .01, ext{prep} > .95 (significant difference, replications of Experiment 2)
  • Both groups compared to fair division ($3): sunglasses vs fair: t(40) = -5.40, p < .001, ext{prep} > .99; clear glasses vs fair: t(41) = -1.06, p = .30, ext{prep} = .65 (only sunglasses are significantly below fair share)
• Perceived anonymity results:
  • Sunglasses: M = 4.73, SD = 1.10
  • Clear glasses: M = 4.01, SD = 1.17
  • Statistical test: t(81) = 2.87, p < .01, ext{prep} > .95 (sunglasses rated as more anonymous)
• Mediation analysis (Baron & Kenny approach, 1986):
  • Path from sunglasses to offers: coefficient eta = -0.29, p < .01, ext{prep} > .95
  • When perceived anonymity is included as mediator, the direct effect drops to non-significance: eta = -0.09, p = .28, ext{prep} = .66
  • Indirect effect: significant via perceived anonymity; bootstrap 99% bias-corrected CI excludes zero: CI_{99 ext{%}} = [-0.77, -0.75] from MacKinnon, Fairchild, & Fritz (2007).
• Interpretation:
  • Perceived anonymity mediates the effect of darkness on the amount offered in the dictator game, supporting the idea that darkness licenses selfish behavior through subjective concealment.
• Table 1: Items used to measure perceived anonymity and concealed identity (5 items, with items 1, 2, 4, 5 marked as not reversed and item 3 as reverse-scored or standard wording per table notes).
• Additional notes:
  • 1) No participant underreported performance (Experiment 1 note).
  • 2) Glasses used did not have prescription lenses.
• Overall takeaway from Experiment 3:
  • Darkness via sunglasses increased perceived anonymity, which in turn reduced generosity in an anonymous interactive task; the indirect effect is statistically robust and supported by bootstrap analyses.

General Discussion and Implications

• Across three experiments, darkness increased dishonesty and self-interested behavior, even when actual anonymity was present or guaranteed.
• The mechanism is illusory anonymity: the phenomenological experience of darkness leads people to feel concealed and less observable, prompting riskier or less prosocial choices.
• Darkness did not alter actual anonymity in any experiment:
  • Experiment 1: participants could still be seen by others; the room was dim but not opaque.
  • Experiments 2 and 3: online, computer-mediated tasks with no face-to-face interactions; sunglasses did not reduce actual ability to be observed.
• The results align with and extend several literatures:
  • Illusion of transparency (Gilovich, Savitsky, & Medvec, 1998): people overestimate how much others can read their internal states; darkness exacerbates this illusion by reducing perceived visibility.
  • Deindividuation and anonymity research (e.g., Festinger, Pepitone, & Newcomb, 1952; Zimbardo, 1969): darkness interacts with self-perception to license behavior that defies social norms; however, this paper emphasizes a phenomenological route rather than purely situational cues.
  • Egocentrism in perspective-taking (Piaget, 1936; Piaget & Inhelder, 1956; Epley, Morewedge, & Keysar, 2004): adults remain partly egocentric; darkness amplifies the anchoring on one’s own experience and reduces attention to others’ perspectives.
• Policy and real-world relevance:
  • Urban lighting and street design: ensuring adequate lighting could reduce dishonest or self-interested acts in public spaces by limiting illusory anonymity.
  • Design of online interactions: even in online or mediated contexts, visual cues that imply concealment can affect generosity and fairness.
  • Ethical and safety considerations: the potential harms of darkness-induced illusory anonymity (e.g., in crowds, online communities, or situations involving moral choices) warrant attention.
• Limitations and future directions:
  • Population: all experiments used undergraduate students; generalizability to broader or cross-cultural populations remains to be tested.
  • Ecological validity: while the lab tasks isolate specific mechanisms, real-world settings involve more complex social cues and incentives.
  • Measurement of anonymity: the mediation relies on perceived anonymity scales; future work could integrate behavioral proxies for perceived visibility or actual exposure.
• Conclusion:
  • The saying “good lamps are the best police” captures a real psychological phenomenon: darkness can trigger illusory anonymity that licenses dishonest and self-interested behavior, even when people believe they are fully anonymous. This effect operates through a subjective sense of concealment and has robust empirical support across multiple experimental paradigms.

Formulas, indices, and key numerical references (summary)

• Difference score for Experiment 1 cheating
  • D = ext{Self-reported correctly solved} - ext{Actual correctly solved}
• Experiment 1: actual performance comparison
  • ext{Actual: } M{ ext{dim}} = 7.26,\, SD = 2.27;\ M{ ext{control}} = 6.95,\, SD = 2.49
  • t(82) < 1,\, p = .56,\, ext{prep} = .46
• Experiment 1: self-reported performance
  • M{ ext{dim}} = 11.47,\, SD = 4.32;\ M{ ext{control}} = 7.78,\, SD = 3.09
  • t(82) = 4.48,\, p < .001,\, ext{prep} > .99
• Experiment 1: cheating indicators
  • Overstatement amount: M{ ext{dim}} = 4.21,\, SD = 4.12;\ M{ ext{control}} = 0.83,\, SD = 1.58
  • t(82) = 4.92,\, p < .001,\, ext{prep} > .99
  • Overstated proportion: 60.5 ext{ extpercent} ext{ vs } 24.4 ext{ extpercent}
  • ext{ } \chi^2(1, N = 84) = 11.15,\, p = .001,\, ext{prep} = .99
• Experiment 2: dictator game offers
  • Sunglasses: M = 1.81, ext{ SD} = 1.30
  • Clear glasses: M = 2.71, ext{ SD} = 1.83
  • t(48) = 2.02,\, p = .049,\, ext{prep} = .88
• Experiment 2 vs fair division
  • Sunglasses vs fair: t(25) = -4.688,\, p < .01,\, ext{prep} > .95
  • Clear glasses vs fair: t(23) = 0.78,\, p = .44,\, ext{prep} = .54
• Experiment 3: dictator offers
  • Sunglasses: M = 1.93,\, SD = 1.27
  • Clear glasses: M = 2.76,\, SD = 1.46
  • t(81) = -2.77,\, p < .01,\, ext{prep} > .95
• Experiment 3: perceived anonymity
  • Sunglasses: M = 4.73,\, SD = 1.10
  • Clear glasses: M = 4.01,\, SD = 1.17
  • t(81) = 2.87,\, p < .01,\, ext{prep} > .95
• Mediation in Experiment 3
  • Direct path: eta_{ ext{Sunglasses} o ext{Offer}} = -0.29,\, p < .01,\, ext{prep} > .95
  • With mediator: eta = -0.09,\, p = .28,\, ext{prep} = .66 (direct effect reduces to non-significance)
  • Indirect effect: 99% BC CI: CI_{99 ext{%}} = [-0.77, -0.75]
• Reliability: perceived anonymity scale: ext{Cronbach's } ackslash alpha = .93

References and theoretical anchors cited in the notes

• Plato’s Ring of Gyges (philosophical precedent for invisibility and corruption) and related discussions in the text.
• Zimbardo (1969) on deindividuation, including the broader social psychology literature on identity, attention, and control.
• Baron & Kenny (1986) mediation model and the classic mediator-moderator distinction.
• MacKinnon, Fairchild, & Fritz (2007) mediation analysis approaches and bootstrapping confidence intervals for indirect effects.
• Other foundational works cited include: Prentice-Dunn & Rogers (1980); Festinger, Pepitone, & Newcomb (1952); Singer, Brush, & Lublin (1965); Epley, Morewedge, & Keysar (2004); Gilovich, Savitsky, & Medvec (1998); Tversky & Kahneman (1974); Piaget (1936); Piaget & Inhelder (1956); Mazar, Amir, & Ariely (2008); Gino, Ayal, & Ariely (2009); Hartley (1974); Bouman (1987).

Takeaways for exam preparation

• Darkness can increase dishonest and self-interested behavior via illusory anonymity, even when actual anonymity is present.
• The mechanism was demonstrated through three experiments using both room dimness and sunglasses to induce illusory anonymity.
• Perceived anonymity mediates the darkness-to-behavior link (Experiment 3), with robust indirect effects evidenced by bootstrap confidence intervals.
• The findings integrate and extend theories about deindividuation, egocentrism, and the illusion of transparency, with real-world implications for lighting design, online interactions, and moral decision-making.