Financial Accounting Standards Research Initiative

I have fairly strong feelings against the use of overt deception (i.e., willful, explicit false statements) to research participants. First, I do not buy the argument that “any potential negative effects on partipants will be removed after the researcher informs them of the deception.” If your spouse or significant other were unfaithful to you but informed you afterwards, would that remove the harm of the infidelity? Admittedly that analogy is extreme, but in a much milder sense, I think we have a relationship with our research participants, and that relationship demands its own form of faithfulness and integrity. Participants are adult volunteers in our research, not lab rats, and they deserve the same honesty and integrity from us that we expect from them. If we insist on absolute honesty in class exercises, for example, and do not buy student rationalizations about breaching the rules “because it helped me to learn better,” then how can we apply a double standard in how we treat those very same students (assuming student participants) in our research? When we lie, we send a very dangerous signal.

Even setting aside the ethical issues, however, there are good scientific reasons to avoid deception, at least in the variety of research I conduct (”experimental economics”). First, there is the “poison the well” problem. In experimental economics, participants have to believe that the researcher will run the experiment (and compensate participants) exactly as the instructions indicate, or control is lost. Consistent with the “once burned, twice shy” adage, deceived participants are less likely to believe the experimenter the next time, even if the (second) experimenter is playing it straight. And more fundamentally, if an experimenter can get away with deception, is the experimental setting really credible? Experimental economists pride themselves in setting up real economies with real consequences, even if they are in a laboratory setting. Deception undermines that realism.

I understand fully that the traditions of psychology differ sharply on some of these questions. If in my current capacity as senior editor of The Accounting Review I were to receive a experiment following the traditions of judgment and decision-making research in psychology that employed deception, I would likely recuse myself from serving as the editor for that manuscript to avoid letting my tastes and biases play too heavy a role (though if perceived-to-be neutral reviewers recommend rejection due to the deception, I think that’s a legitimate point for the editor to consider in a peer-reviewed community). So I think the experience relayed in the original comment (i.e., a table of attendees at a consortium who were sharply divided on the issue) is fairly typical. This is a very controversial topic in which people take very different viewpoints. I just wanted to express a counterpoint from my perspective as an experimentalist who has adopted a “no deception” pledge for my research. Thanks for listening.

Steven Kachelmeier
University of Texas at Austin

Definitely a thought-provoking post, Corey. I think Steve has captured the two key concerns about deception well: that it creates negative externalities that harm experiments who rely on a subject pool that trusts them, and deception is simply wrong.

My own take on the issue is one of pragmatics, not morals. My first piece of advice would be not to employ deception unless there is no way to address the research question otherwise. I don’t think there are many research questions in accounting that can’t be addressed without deception. Perhaps the next time a table of people are talking about the issue, you can press advocates of deception to describe a study in which they think they don’t have an alternative. That will narrow the discussion quite a bit.

I also think it is a mistake to frame the question as “deception: yes or no.” What is deception? For example, is it deception in a laboratory market to tell subjects that security values are being drawn randomly, but actually you have chosen a predetermined set of random numbers? What if you intentionally insert a security value that is extremely high or low, in order to strengthen a manipulation–while still ensuring that the values seen by participants are indistinguishable from the stated distribution? What if you imply that the study is about how people trade given the information presented to them, but really you are manipulating the room temperature or background music to see how often they are distracted from the task? What if you have confederate traders whom you asked to trade very aggressively? I know Steve takes a pretty hard line on this stuff, but I wonder which of these he would feel are unjustifiable.

To me, the one thing you really want to avoid is explicitly telling people something that is patently false. That is why I say deception can so easily be avoided. For example, instead of telling subjects “we used a random number generator to create security values” you can tell subjects “the security values you see are very similar to what you might see if values were distributed uniformly from 0 to 100.” And if you have confederate traders, don’t say you don’t.

As long as everyone follows this rule, experimenters who don’t engage in deception can still gain the trust of their subjects by being very clear. I will typically tell my subjects “This experiment was approved by Cornell University’s Internal Review Board as one using no deception of any kind, and I can’t tell you that if it isn’t true.”

Internal Review Boards are a pain in the butt, but here is one place they are our friends!

Rob’s examples are excellent illustrations of gray areas in which even I (an admitted hard-liner against deception) would let reason prevail. No, I do not think it would be a problem to predetermine quasi-random sequences, especially if the experimenter informs participants that “the values you see are similar to those that would arise from a uniform 1-to-100 distribution” in lieu of the more problematic “the values you see are being produced by a random number generator” if in fact you predetermined the numbers.

Somewhat ironcially, once in my own research, a reviewer criticized me for deceiving participants by not telling them beforehand that, after 10 periods, they would reverse roles in an interactive game. I never indicated or implied otherwise to participants, and I felt that the role-reversal was pretty innocuous. But this anecdote goes to show just how sensitive people can get about these things. To be sure, I think it is deception if you are selling your used car and don’t happen to mention to the prospective buyer that it uses a quart of oil every 50 miles, even if you don’t represent otherwise. But that doesn’t mean we have to reveal all our hypotheses and design subtleties. Sound judgment and a sense of research ethics trumps “rules” every time.

So in the end, it boils down to a judgment call. Rob’s examples are excellent illustrations of where any view (including mine) can lead to silly consequences if taken to the extreme. So just ask yourself if you would feel deceived, and if so, if that would bother you if you were a participant under the circumstances. If the answer is “yes,” I advise caution. Oh, what a tangled web we weave ….

Steve

Corey mentions in the original post that

Later in the day, the question was raised in a more formal setting, and a well respected presenter at the consortium did not hesitate to say that he had absolutely no problem with this type of research as long as the participants were given the truth immediately after the experiment was performed.

Who was the ‘well respected presenter’? Presumably this isn’t classified information, since the comment was public. Maybe we could get them to comment on the post as well.

I completely agree with the comments made by both Steve and Rob. Steve clearly explained why it is very dangerous to use deception in research and Rob’s examples are great illustrations that the issue is not always black and white. Steve’s rule of thumb to consider whether you would feel deceived and whether you would be bothered by that is a good barometer to gauge the effect that the experiment may have on participants. In the end, using deception in experiments carries the risk of diluting the perceived validity of our research. Just as the value of an audit depends on the body of auditors maintaining independence, the value of experimental research will be harmed if experiments gain a reputation that they may involve deception.

I appreciate and have enjoyed reading your comments. I will invite the presenter I mentioned in the previous blog to join in the discussion (I hesitate to reveal his name without contacting him first). While there often seems to be some division on the matter, I find it interesting that those who are opposed to deception in research are usually strongly opposed, while those who are not necessarily opposed really don’t seem to have strong feelings in favor of using deception. That, in and of itself, is very revealing on the matter. It appears that any researcher who wishes to use deception will be faced with the task of justifying such use to the research community, and perhaps there is no acceptable justification.