Explanation matters. Explanation is a central part of coming to an understanding of the world around us. So it is no surprise that the question of what makes “a good explanation” has been a long-standing interest of philosophers, in particular philosophers of science. Their work, in turn, has influenced and motivated psychologists interested in everyday explanation.
But explanation is not just a feature of interest-free inquiry, where we simply try to understand the world as it presents itself to us. Explanation also figures in our attempts to convince others: In arguing for a particular economic policy, for example, we might appeal to an explanation of how such a policy will be effective; or, looking back in time, we might want to persuade others to avoid mistakes of the past by appealing to an explanation of why a particular policy had ill-effects. In other words, explanation has close ties to argumentation and persuasion.
Explanation also has close links with what philosophers refer to as the problem of testimony: as Zemla and colleagues in the special issue on explanation point out, we frequently rely on explanations provided to us by others, and those explanations may concern things of which we have little or no first-hand knowledge or understanding. In this situation, we must make judgments both about how convincing an explanation seems (given only limited information against which to check it) and how trustworthy we consider the source providing it to be. In short, explanation in real-world contexts will often involve the interaction of a broad range of factors and concerns. And many of these have not been the focus of theorising about “good explanations” in the context of science.
Some of these factors, however, have become a societal focus in recent years, as illustrated by the debate about the role of experts in the US or the UK, Commentators such as Thomas Nichols have (rightly or wrongly) bemoaned “the death of expertise”, and high ranking politicians have proclaimed that “we have all had enough of experts….” (former UK education minister Michael Gove, 2016).
All of this makes it exciting that the article by Zemla and colleagues takes as its point of departure real-world examples of explanation retrieved from Reddit’s “Explain Like I’m 5”, an internet community that receives several million visitors every month. Zemla and colleagues selected from that repository sample explanations from a range of domains (historical, public health related, legal, and concerning social policy). Participants rated not only the overall quality of the explanations but a list of specific attributes that one might expect to be relevant to explanation quality. These ranged from central “explanatory virtues” within the philosophy of science such as simplicity, or coherence, through to perceived truth of the explanation, the quality of articulation, novelty, or the likely existence of alternative explanations.
The main finding was that, far from preferring simpler over more complex explanations, there was evidence that participants perceived more complex explanations that invoked multiple causes to be of better quality. This is surprising in as much as the theoretical literature within the philosophy of science has stressed simplicity as a key explanatory virtue. The intuition that, at least in the context of the natural world, simpler explanations are better has long roots. Zemla and colleagues cite Thomas Aquinas the 13th century philosopher and theologian: “nature does not employ two instruments when one suffices.” And this bias toward simplicity is reflected in many things, including statistical criteria for model selection.
At the moment, the reason for this divergence is unclear. However, Zemla and colleagues do confirm it in a follow-up study using controlled experimental materials. Based once again on material found on the reddit site, the experimental follow-up manipulates whether participants received one of two possible causal mechanisms as an explanation or both of these together. Results, again, indicated a preference for complexity, with participants preferring the concatenated explanations that included two causal mechanisms over one.
Another interesting feature of the paper by Zemla and colleagues is that it found evidence for the role of perceived expertise on perceived explanation quality. This may already seem intuitive, but it connects the issue of explanation quality firmly with the problem of testimony. In so doing, it confirms the importance of the wider pragmatic contexts in which everyday argument occurs: contexts that involve goals, not just interest-free inquiry, and that involve specific individuals communicating with one another. This brings explanation together with the realm of argumentation and persuasion.
Deanne Kuhn’s classic study “The Skills of Argument” examined people’s skill at generating arguments in naturalistic contexts, asking them to provide causes and evidence for those causes in three naturalistic contexts similar to those examined by Zemla and colleagues (school performance, criminal re-offending, unemployment). The study probed people’s ability to engage in real world argument, both across their lifespan and across different levels of educational background. They asked people to provide a causal explanation, but also to consider alternative causes for the phenomena in question and to indicate what kind of evidence would help distinguish between these competing causes. Kuhn and colleagues found parallels between the kinds of reasons and evidence people generated and what they accepted as evidence and explanation from others.
Viewed from the perspective of argument, Zemla and colleague’s results seem less surprising: multiple arguments may make a stronger case than just one. And they may do so not just psychologically, but also normatively. In short, considering the wider pragmatic context of explanation in the real world reveals the overlap of explanation with the concerns of argument and testimony. This opens a perspective on a much wider range of normative concerns (that is, evaluative concerns about quality). Broadening that perspective on explanation seems more far-reaching than any individual result in a psychology study is ever likely to be. The investigation by Zemla and colleagues is to be commended for pushing toward such a shift, and researchers from a whole range of areas should be excited by the issues it helps bring into view.