What is an explanation?


Yesterday, I was discussing a book with an anthropologist. What could have been an interesting discussion, became, from my perspective, a futile one. We could not even agree on what constitutes an explanation and how such an explanation should be put to the test. This caused me to think: `What exactly is an explanation?’

To my opinion, the book we discussed is a very interesting and possibly inspiring one, but it does not yet offer an explanation ((NOTE: this is my first and preliminary reaction to a discussion I’ve had recently. It still fascinates me, so I’ll probably find myself writing about it more often in the near future. I will then give more details on the book I mentioned as well.)). Focused on the ‘narcissism of the minor differences‘, it states that conflicts arise especially when differences between groups are very small. I think this is an interesting thesis, but not yet an explanation. My main reason for this is that it does not explicate when a difference is small (as opposed to large) and what exactly is a conflict. According to my debating-partner, all can be interpreted as conflict and in every context it differs what is a ‘minor’ difference. Thereby, the supposed `explanation’ does not lead to new insights, for every ‘explained’ situation is described by even so much determinants. To me, that is not an explanation, but a label on a correlation.

Then the testing issue: the book illustrates the thesis of the minor differences nicely. This helps the reader to understand the arguments, and shows the potential value of it. But can we regard this as a test of the explanation? I think not. A proper test as I would like to see it compares the prediction / explanation under different circumstances, so not only the successes. If indeed minor differences between (the people in) groups lead to conflicts, than we should also be able to understand why in other situations no conflict arises when no small differences are present, or why when small differences are present not in all situations conflicts arise. All these types of situations should be compared with empirical reality.

So what it comes down to, is two things. I do not think that labeling a correlation is the same as offering an explanation. And I do not think that illustrating an explanation is the same as testing an explanation. Also, I fail to understand how empirical science could possibly make any progress when these arguments are not valid. Do you have any suggestions?

4 comment on “What is an explanation?

  • No, I have no suggestions because you are spot on. A correlation is just that, a correlation. However, that doesn’t mean that the correlation might be wrong. It is at that point that we start to explain the correlation.

    The problem is, I think, that in this case the correlation can’t even become an explanation because there aren’t enough factors determining how the correlation came to be seen in the first place. I hope I guess correctly when I say, that they probably took whatever they needed from the empirical world to show the correlation, without putting them in context. The context being, that conflict can probably arise when differences are very big too and, as you said, may not arise when the differences are very small. If you have enough examples of similar instances, like small differences leading to conflict, then this doesn’t prove that small differences between groups leads to conflict.

    I know, this is like pouring salt into an open wound. But these kind of supposed explanations ask for such obvious complaints. Hopefully, in some distant future, anthropologists will understand the difference between a correlation and an explanation too. That’s all I can say about the subject.

  • If you/we are in the mood of fundamental discussions, I would wonder first what do you mean by empirical science. Opposed to what? is out there “non-empirical science”?

    I raise the question because ultimately, if you become real popperian, it is not possible to do science regarding social events. You can never control for all variables. So you must allow some level of non-objective decision, for example regarding what is big and what is small at evaluating a real social event.

    So, even if I in general agree that the goal of being as objective as possible is desirable, I would be more or less tolerant of alternative definitions of what an explanation is…

  • But an explanation has to exlain something. Or, rather, something has to be explained by an explanation. I think that that is what is missing in the story described above. Empirical science itself has nothing to do with it. The fact that Rense can’t see a future for empirical science is a cry for sanity concerning something as simple as the meaning of what an explanation is.

    But that isn’t really what this is al about, empirical science. I don’t think the problem concerns something like an objective truth. It is just that saying things like ‘small differences lead to conflict’ is not a explanation for the fact that small differences lead to conflict. That is like saying that ever time we eat icecream, we get a cold tongue without saying that it is so, because ice has a temparature below zero, which is far below the normal temperature of a human tongue… Etc..

    Of course, we can find situations where small differences lead to conflict between groups, but as I said, because there might not be and conflict between other groups with small differences between them, this doesn’t explain anything. The most we can say is that there seems to be a correlation, which isn’t the same as an explanation.

  • Dear Inti,

    as you know, I’m always in the mood for a fundamental discussion 😉 The reason that I explicitly mentioned empirical science, is not to get into the debate whether or not disciplines as philosophy are to be regarded as science.

    I do think that it is possible to do observations that are interesting for social studies. Indeed, the level of accuracy is lower than in physical science (however, on what scale?) and it is not possible to do real experiments, controlling for all variables. However, it is possible to control all conceivable variables. Thereby, I do think that many of the social science articles I read are indeed testing hypotheses (and sometimes rejecting them) in a Popperian sense. Perhaps the level of uncertainty is indeed larger than with sciences that do experiment.

    But, I do not think that the discussion regarding what an explanation is has anything to do with observations. As Jasper points out, I’m trying to make a logical argument, not one regarding the validity of observations. Science proposes explanations, which are hypothetical in nature and empirically tested. I do not think that we should be more or less tolerant regarding the issue what the nature of a proper explanation should be. The discussion regarding tolerance is only interesting when it comes to testing the explanations.

Leave a Reply to Rense Nieuwenhuis Cancel reply