Forgetting past misdeeds to justify future ones

To find out, a team of behavioural economists from the CNRS (1) recruited 1322 volunteers in an online experiment which took place over two sessions. The first session involved 20 repetitions of a lottery task, the results of which determined the participants' monetary payoff; however, as the participants had to self-report the outcomes they had the opportunity to cheat (2).

During the second session, three weeks later, the same participants were monetarily incentivised to recall as accurately as possible the outcomes they had reported in the previous session. Half of the volunteers were informed that they would then have the opportunity to voluntarily return some of the money if they had overreported their outcomes in the first session.

It was within this configuration that the participants were most prone to amnesia, as reported by scientists in PNAS on 28th September 2020. In other words, they remembered their cheating behaviour less accurately when they knew they would have to make a moral decision again, even though they could earn more money by remembering the reported outcomes.

It was as if forgetting this incident would allow them to restore their reputation, making it more acceptable for a future breach of morality.

Notes:

(1) Working at the Groupe d'analyse et de théorie économique Lyon -- Saint-Etienne (CNRS/Université Claude Bernard Lyon 1/Université Lumière Lyon 2/Université Jean Monnet/ENS de Lyon) and at the Sciences économiques -- Sciences Po laboratory (CNRS/Sciences Po).

(2) Using a mind game, cheating could not be detected directly by the experimenters. Scientists used statistics to estimate cheating behaviour at an individual level by asking the volunteers to play 20 rounds of a "wheel game" where numbers, between 1 and 6, were randomly displayed. Any results that deviated too far from the expected distribution (more 5's and 6's than pure chance would have allowed) were considered to be due to cheating. A participant was thus classified as dishonest if the average of reported numbers was greater than 3.5 (the expected average for honest reports) and the distribution of reported numbers was significantly different from the uniform distribution.