People Are More Likely to Cheat at the End | Hacker News Reader

People Are More Likely to Cheat at the End

scientificamerican.com

82 pointsdnt404-112 comments

Towards the end of an iterated prisoner's dilemma, your opponent has fewer upcoming opportunities to punish you, and so you may as well defect (especially since your opponent may be thinking the same thing). The interaction in the study really doesn't resemble an iterated prisoner's dilemma, but this may still explain why people instinctively defect more as an ongoing interaction is nearing completion.

The interaction resembles a repeated PD because people are cheating the experimenter :D

The ending effect happens all the time in experiments. If people play the repeated prisoner's dilemma, they'd cooperate UNTIL the last round. If people play the public good game, they'd contribute generously UNTIL the last round.

Because if the prioner's dilemma and public good game are played one shot, the rational thing to do is to cheat right away (defect / not contribute). Being cooperative (in hope to make the world a better place) goes against the selfish motives.

So, at the last round, because there is no tomorrow, it's just like the game is played one-shot, people'd cheat. The theory of game goes too far to say that, if people are rational, they work the game backward (backward induction) and at the 9th round, they'd cheat, and at the 8th round, they'd cheat...-> they'd cheat right from the first round.

But experiments show that people are myopic or at least dont cheat right in the first round, they wait until last round. Because of experiments, theorists have to add the social image. Social image can be built in the first 9 round, but in the last round, not anymore.

It's fascinating the number of plausible arguments people come up with for this behavior.

One that occurs to me if that someone has unlimited tosses, then the number of correct guesses doesn't matter that much for their "expected income stream" - if they want more money, they can just toss more coins instead of cheating. And oppositely, if the experiment is coming to an end, the only way to get more money is by cheating and so now the incentive to cheat becomes a factor.

This isn't fundamentally that different than kazinator's approach [1] but it is a little different.

The position I take on all this is that psychological experiment always involve some particular, reproducible series of behaviors that are then given an abstract intuitive interpretation (distorting heads-tails results becomes "cheating-in-the-abstract"). But naturally any particular behavior actually can have legion abstract interpretations and explanation. I suspect only a much large collection of behaviors and abstraction interpretations could serve as a real compelling explanation. The problem is the intuitive human "toolbox" of abstract explanation - such as "cheating"(in the abstract) might not be the real toolbox we need to use to explain human behavior.

[1]https://news.ycombinator.com/item?id=10223289

Are we "owed" money beyond what we earned from the coin flips or essay grading? Maybe not but perhaps we feel that way. One analysis of the "wasted employee time" due to slacking off with solitaire and the like - clearly this is an old study - found it was basically a way for employees to only do the amount of work they felt they were being paid for. A little way of getting back at the employer for being, as they felt, underpaid.

Which is to say, I wonder if this effect might change if the participants were offered more money?

Or autonomy: tell participants they need to design their own coin-flipping experiment, then perform, measure, and report the results. They still might be underpaid on a $/time basis, but with a greater investment in the process I suspect they'd have a certain amount of pride in ensuring accurate results.

People are likely to cheat toward the end for a different obvious reason: namely that their estimate of impending poor success more and more accurate as the end nears, and so the pressure to do something to improve the outcome increases.

(This is in a different setting, when there is some limited time to achieve some predetermined, measurable goal.)

When the pressure sets in is when people start cutting corners, and that can come simply from being closer to a deadline while lagging in progress.

Really interesting research. I definitely felt something similar when I came to the end of a job recently, after three years in post. I didn't sabotage anything but I did relax a little more and maybe I wasn't as productive as I had been previously, which you could say was 'cheating' slightly.

That is almost exactly what Daniel Ariely's research on dishonesty is about, this notion of cheating so slightly you don't feel guilty about it:

https://www.youtube.com/watch?v=XBmJay_qdNc

> This is reassuring for those optimists out there, as it suggests that people are often honest even when they don’t have to be.

Or it could be that people don't realize there's no way for the experimenters to know if they're cheating, so they're on their best behavior until they become familiar with the system.

I think this is true in a lot of domains (jobs, dating, etc). In the beginning, people follow the rules. When they're comfortable with the ins and out of the system, they start gaming it.

You didn't read the article.

The more time you are to remain in a context, the higher the risk of being punished.

This is probably the reason.

Reply on news.ycombinator.com