The experiment:

People are asked to guess the outcome of a coin flip. If they guess correctly, they get a reward. However, people only reveal their initial guess after they've seen the outcome of the coin flip. This allows them the possibility to cheat, by lying about their guess.

Hence, we can measure dishonesty by seeing whether the proportion of correct guesses is significantly greater than 50 %.

My question:

Does this study really measure dishonesty? Does it not rather measure carelessness? Here's what I mean. Unless you happen to be a complete idiot, if you are invited to such a study, you know what the point of it is. Everyone with a brain can figure out it's an experiment to test your honesty.

Hence, this knowledge will possibly impact their actions during the experiment. For example, if you are a person who cares a lot about your reputation and perception amongst other people, you might think to yourself

oh no, if I lie about my guess, the experiment might conclude that I'm a dishonest person. I don't want that, so let me just play it honestly.

Now, such a person might very well be extremely dishonest ordinarily, but since they are now in a situation where they know their dishonesty is being measured, they'll try to act honestly.

So in this case, the study would conclude such a person is honest, when in fact they are the quite opposite.

Likewise, an otherwise honest person might look at this experiment and think to themselves

I'm a honest person in life ... but this is just a silly, little game, isn't it? Let me just cheat, it'll be fun and I'll get a reward.

This person is honest, but is just having a little fun with the experiment. Yet the experiment would conclude that this person is actually the unethical one.

Am I right? If so, what does that say about all studies that attempt to measure people's ethics. Don't they all suffer from this phenomenon where the bad people know they are being "watched" and try to act good, while the good people don't really care whether or not they are being watched, since they aren't that insecure about themselves and don't feel the need to convince others they are good?

And if so, what other type of experiment would actually be capable of measuring somebody's ethics without committing these fallacies?

4 Answers 4


Yes, your example is silly. Someone who took that experiment seriously would have a big problem defending it and would probably not get published.

But there are well-planned and adequately subtle ethical experiments. You just have to lie to the subjects. You have to reasonably convince them that you are studying something other than ethical behavior. And the convincing has to be adequately realistic.

For example, a lot of psychology experiments involve competitive games, and a reasonable number of ethical experiments simply repurpose one of those, so that it looks like a replication of the existing study, but track things like taking opportunities to cheat, instead of using the game as it was originally intended.


In ordinary life, people have many incentives for being honest (that includes being watched by other people or by god), and dishonest(various rewards).

And we all perform a balancing act when we choose how to behave.

In this experiement, as you report it, dishonesty is incentivized by the reward, but honesty isnt much incentivised (and isnt for people who ignore probability laws, there are A LOT of such people among gamblers). Therefore the results can be extrapoled for similar life situations.

Each experiment covers a given situation, no experiment covers them all. I suggest you take a look at a similar experiment with bagels , you may also find a good summary of the experiment in freakonomics book.


Does this study really measure dishonesty?

No. This test is too simplistic and unable to distinguish between honesty and genuine faculties such as intuition or organized thinking. A person may opt to compute his initial guesses through an effective algorithm and truthfully disclose his result each time.

In other settings such as discovery in court proceedings, a battery of psychological tests is applied to detect malingering. Since intent is a prima facie element to be proved in many offenses and torts, guilty defendants oftentimes exaggerate their mental state to convince the judge/jury that they did not intend to cause unlawful harm. Likewise, a plaintiff who seeks recovery for intentional infliction of emotional distress may be asked to take such tests to ascertain whether or not the defendant's wrongful conduct truly is the cause (whether proximate or not) of his distress.

Even a non-expert can unearth evidence of malingering. Few years ago I sued my former employer for defamation. To elude an adverse judgment, he insisted that he interpreted as "direct threats" a proverb I wrote in two emails. When I finally was able to take his deposition (for he avoided the deposition for as long as he could), I asked him several consecutive times to articulate his interpretation of that proverb. His answers where:

  • I don't recall
  • I don’t know
  • [I] [e]xplain it to a kid at the age that the kid can understand and comprehend that phrase
  • Right now I don’t have a kid here so I don’t know
  • You are not a kid

(the entire transcript of his deposition is available here in the bullet point for April 18, 2016)

This defendant's inability and clear refusal to articulate his alleged interpretation of a proverb contradicts that he felt threatened by it. Otherwise he would have been eager to testify as to his mental state, more so when he eluded his deposition for three months during which he had the chance to organize his thoughts.

The point I am trying to make is that tests for (dis-)honesty often involve more complexity than a survey of unverifiable data.


Welcome, Maroo, interesting question. However, I don't see how the experiment, or its result, proves dishonesty. After all, every participant might be a truth-teller and totally honest and yet the proportion of correct guesses could still be significantly greater than 50 %. Maybe significantly more participants really did guess correctly.

One may have one's doubts about the honesty of (some of) the participants but their dishonesty is certainly not proved by there being significantly more than 50% who reported having guessed the right outcome.

You add extra assumptions which may not apply and which you do not justify but only assert :

Everyone with a brain can figure out it's an experiment to test your honesty.

Hence, this knowledge will possibly impact their actions during the experiment.

You are - hope this doesn't sound too harsh - mixing statistical probability with unargued empirical assumptions. But you will find your way in philosophy and I hope you continue to contribute to the site.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .