mardi 1 octobre 2013

A few thoughts on the role of paradoxes in social psychology



Source: René Magritte, L'Empire des Lumières. 

Every year, students following their third year in psychology and educational sciences at the university of Brussels have to prepare a "project" loosely related to a set theme, common to all students. The theme must be very broad so that it can be approached in all subfields of these disciplines. This year, it is "paradoxes". I was asked by the coordinators of this activity to deliver a short lecture on what my subfield (social psychology) has to say about this.

After scratching my head a few times, the importance of paradoxes in the development of psychological science in general, and social psychology especially, became quickly obvious. Below is the outcome of my (preliminary) musings on this "set piece".


Etymologically, paradoxes refer to what defies common sense (para/against, doxa/opinion). With this definition in mind, it is not difficult to come up with a host of paradoxical research findings. Actually, when appraising research in social psychology, it is plausible that more reported results are "paradoxical" in this sense than conforming to common sense.

Here are some examples of classical and more recent effects reported by social psychologists: 

These findings (as many others) violate common sense in various ways: by reversing cause and effect, by suggesting that a factor that should have no effect actually has one, by modifying the direction of an "obvious" effect, by showing that a very subtle manipulation can have long lasting effects, etc....

The french philosopher Gaston Bachelard (1938) argued that science was inherently opposed to "opinion" (i.e., doxa). "Opinion thinks wrongly, it does not think, it merely translates needs (...) One cannot build anything on opinion, one must destroy it first " (p. 14, my translation). In social psychology, findings that contradict common sense are more likely to be published because they are deemed more interesting by the research community. Indeed, findings that confirm common sense expectations (what McGuire (1997) called "bubbapsychology") or that do not successfully reject the null hypothesis, are often considered as unworthy of further attention. Paradoxical findings play a crucial role in the development of novel theories of human behavior. And many of the examples cited above actually have generated rich and productive research programmes. The curiosity aroused by these paradoxes played an important role in this development. 

Paradoxical findings are one of the main sources of interest for social psychological research among nonspecialists. When I teach my social cognition class for example, not only do I feel that my students are most interested in the paradoxical and surprising findings but it is much more exciting to describe them. One feels a bit like a storyteller revealing the name of the murderer no one had suspected.  

The interest generated by such findings does not only influence teaching but has consequences on their likelihood of publication. Given that access to publication is extremely competitive (in social psychology journals, the rejection rate of most journals lies between 70% and 90%), and that publication is essential to survival in academia ("publish or perish"), this interest may offer an an edge in publication. In other words, the likelihood that a finding is published may be higher for a paradoxical finding than for a "normal" finding. This may affect all stages of the publication process:

  • researchers may be more likely to run studies that test paradoxical hypotheses
  • researchers may be more likely to report these studies
  • researchers may be more likely to submit these studies to scientific journals
  • scientific journals may be more likely to accept articles presenting these studies

This presents at least two problems.

Problem 1: Validity

One has to do with the validity of published research. Ioannadis infamously concluded from simulation studies that most published research findings are likely to be false (he focused especially on the medical sciences but his argument can be applied to psychology as well) . This is because of several weaknesses (e.g., lack of power, preference for expected findings,...) that lead to the publication of studies (falsely) rejecting the null hypothesis.

To survive, lay beliefs (i.e., common sense), generally have to be empirically validated. Thus, among the many beliefs we could possibly entertain, those that are part of what we call "common sense" are probably more likely to be true than those that aren't.  Hence, paradoxical findings are even more likely to be false ( i.e., that their conclusions cannot be generalized to the overall population) than expected, nonparadoxical findings. 

Several responses can be proposed: replicate the research finding, use statistical methods that take into account the prior probability of a finding (i.e., bayesian methods), and provide a compelling and testable theory accounting for it (which has been done for some of the findings mentioned above). Although this is what should be done, as there are yet very few incentives for replication, it is not obvious that such advice will be followed. This argument  about validity, or variations thereof, has been proposed by several colleagues (see e.g., this, this or this).

Problem 2: Legitimacy

The second problem has to do with the role of paradoxical findings in the legitimization of  psychology, and especially social psychology. Paradoxical findings serve to  buttress psychologists' position as experts. Such findings suggest that, through their methodological and theoretical expertise, psychological scientists have access to a province of knowledge that is inaccessible to the layman. Consider, for example, an excellent book written by distinguished psychological scientists, that questions 50 great myths of popular psychology. Its subtitle is "shattering widespread misconception about human behavior", thereby highlighting the superiority of rigorous psychological science over lay beliefs and "pseudoscience".  Although social psychologists are often very critical of psychoanalysts, they may seek a similar form of legitimacy. Whereas psychoanalysis purport to have access to the mysteries of the unconscious, social psychologists can breach common knowledge and illuminate the true roots of human behavior (I am not sure this wording is felicitous but, please, excuse my less than perfect english!). As the public is particularly thirsty for such findings,  the media may reinforce this trend and, thereby, the legitimacy of the discipline. According to this logic, engaging in systematic replications, many of which find little support for the original findings, may not be a very popular strategy as it may undermine the status of the discipline. 

Pursuing this argument, our predilection for, and the overrepresentation of, paradoxical findings in the literature may promote an undemocratic view of scientific knowledge. Indeed, to some extent, much of social psychological research involves discrediting laymen's discourse. For example, since Nisbett and Wilson's seminal paper, generations of graduate students have been taught to believe that people can't account for the reasons driving their judgments and actions. If so, how can they be responsible citizens? And why should we listen to what they have to say about why they do what they do? (and indeed a host of studies suggests that behaviors that are central to democracy, such as voting, are influenced by factors that people would be hard pressed to acknowledge such as the candidates' facial features). Regardless of the validity of such findings, it is not absurd to wonder whether, in advertising a (possibly paradoxical) view of the citizen of modern democracies as a constant victim of unconscious biases, social psychologists are undermining democratic values.  

The strength of common sense

By emphasizing the limitations of common sense, we may have overlooked its strengths: When scientists try to map common sense with reality, it is often not very far off the mark, probably because what we call "common sense" has to be somewhat empirically validated to survive. For example, Lee Jussim and his team have for years tried to estimate the accuracy of social stereotypes, one of the prime examples of "common sense". In one study they sought to compare teachers' estimations of the impact of variables such as gender, social class and ethnicity on math achievement with the actual effect of these variables on achievement. This may be a questionable endeavor from an epistemological point of view (is it meaningful to consider stereotypes as "true" or "false"?). But, if anything, Jussim's findings suggest that teachers are quite good at estimating group differences. This does not mean that they use these inferences to evaluate individuals or to explain their behavior. Other lay beliefs about psychological functioning may be more difficult to validate empirically and more resistant, even when they are ultimately false. Nonetheless, if anyone dared to count them, I would predict that only a minority could be clearly established to be false. 

Conclusion

In sum, paradoxes benefit (social) psychology in many ways as they play an important role in driving curiosity and stimulating theoretical innovation. Yet, they may be overrepresented in the scientific literature due in part to publication bias. Encouraging the publication of such findings may also serve to legitimize the social position of scholars (such as myself!) working in this discipline.

It may be time to consider studies not in terms of their results (whether they confirm "paradoxical hypotheses" or not). Even if the adequacy of their methodology and conclusions with regard to the hypotheses they put forward should obviously remain a central concern, this does not mean that we should overlook theoretical novelty in judging the quality of an empirical article. But the theoretical value of a paper should not only be judged in terms of the "paradoxicality" of the findings it may predict but also in terms of its articulation with other existing theories and the generality of its predictions.

Finally, given that debunking common sense has become a new doxa, it may be time to question this paradox and appreciate the strengths of common sense as well. 

Reference: 

Bachelard, G. (1938). La Formation de l'esprit scientifique, Gaston Bachelard, éd. Vrin, 1938, p. 14
McGuire, W. J. (1997). Going beyond the banalities of bubbapsychology: A perspectivist social psychology. 
The message of social psychology: Perspectives on mind in society, 221-237.
PS: Thanks to my colleague Christian Orange, for the Bachelard reference.




4 commentaires:

Dave Nussbaum a dit…

Very interesting points, Olivier! I think you hit on some important issues -- paradoxical findings can be critical but have spun out of control.

There's something incredibly important about a counter-intuitive finding when it causes us to have to rethink the way we once thought of things and reveals new insights. Whether that be dissonance findings poking holes in behaviorism, or the white bear finding shedding light on the importance of metacognition.

The problem becomes, as you make clear, when we start looking for counter-intuitive findings for their own sake, rather than what they teach us about our pre-existing beliefs. As McGuire put it, roughly, every hypothesis is true, the question is when and why.

Mining for counter-intuitive hypotheses is not only a problem for the excellent reasons you point out, but also because it undermines the high value of those counter-intuitive findings that are truly important.

Olivier Klein a dit…

Thanks for your comment, Dave. It's nice to see I am not the only admirer of McGuire's perspectivist approach. And I totally agree with your last point (I wish I had written it myself!).

Roger Giner-Sorolla a dit…

You can publish the incredibly obvious. There just has to be a picture of a brain thinking it.

Olivier Klein a dit…

I guess you are right, Roger. I read too little (social) neuroscience.