There have been many attempts to define culture, each having its own spin reflecting what work the definition needs to do. My favorites are Boyd & Richerson's and Page & Bednar's, though they are quite different from each other. What the existing definitions have in common is that they are attempts to capture what culture is rather than what culture does. My claim is that culture is properly understood as systematic patterns in how things are done; and my focus is on measuring cultural distances rather than simply identifying cultural components or providing culture-based explanations. My measurement scheme analyses similarities in the differences of how things are done: the "same difference" criterion. This measurement does not help provide an explanation for culture and their differences, but it does identify which behavioral features need a cultural explanation. A sketch of the technique follows.

I first came across this measuring concept while trying to find a way to discover anomalous behavior. It then occurred to me that behavior might be anomalous in domain X because it is adaptive for domain Y but happening in domain X. The domain might be occupation, geographic region, time of day, or really anything that can have a systematic effect on the behavior under investigation. So, for example, if somebody is bowing as a greeting in America then that would be rather anomalous for America. Assume that we have a stock of models for greeting behaviors. While observing the behavior we could therefore 1) identify it as Japanese greeting behavior, 2) recognize that it is not happening in Japan, and 3) tag it as anomalous. That's fine for what it is, but that's not culture differentiating because it assumes we already have the models for each behavior.

To discover culture differences we need more data points. The first step is to observe a statistically sufficiently large number of greeting events in a variety of contexts in both domains, let's say greetings at home among family, greetings at work among colleagues, and greetings in a café among friends - all in both America and Japan. That's six categories of events in a 2 x 3 matrix: two domains and three locations. The next step is to find the commonalities of behavior at each location, e.g. what qualities are shared for greetings at work regardless of domain. Those qualities are the ones that are likely practical requirements for some event to be a greeting (though there may be some shared cultural features in there too). The remainder of the events' qualities after subtracting the common ones is the set of qualities requiring addition explanation, part of which might be a cultural explanation. We perform this analysis for each location so we end up with three groups of unshared qualities for each domain.

The next step is to compare the resulting unshared properties across the locations for each domain. The qualities that all the events share that cannot be explained by similarities in their practical requirements are then ripe for explanation as cultural features. So the methodology here is to find differences in how some activities are performed and then to determine which differences are the same for multiple activities within a culture. That's why it has the "same difference" moniker. Conceptually that's the story and I hope that's convincing as a reasonable measure of the sorts of features that cultural explanation is expected to elucidate. Specifically I want this procedure to capture those differences that we would identify as cultural differences. Now comes the hard part: quantification.

Even if the above characterization is accepted, it's not clear that if would be very beneficial for analyzing narrative accounts of behavior, but I don't really spend much time around such things so I don't know (if you know then please email me and tell me). It might provide a conceptual boost for further thinking at least. Even a qualitative representation of the features of the behaviors in each domain would still leave a gap in one's ability to determine the degree of cultural distance, but certainly a binary distance measure (i.e. sum over {same = 1, different = 0} for all features) could be attempted. But if the behaviors under consideration admit to a highly parameterized representation then there are all sorts of techniques for grouping and comparing sets of parameter values. These are the cases I had in mind when I came up with this methodology, determining cultural distances in agent-based models of societies where behavior is often measurable with clear quantitative metrics.

As for feeding back into definitions of culture and its purported presence in some model, the above methodology also establishes a minimal requirement for culture to be present. For something to be a cultural variant it must be revealed in such a way that within that model's framework a measurably different culture could have arisen. So culture can't just be a consistent behavior, or a coordinating equilibrium, or a shared strategy, but in some models those things could reveal cultural differences. The side of the road on which people drive is a standard example of cultural behavior, but using this definition that feature would only contribute to a culture distance measure if societies that differed in that respect also tended to ride bikes, walk on the sidewalk, have entrances to buildings, etc. on their respective sides as well. I'm making the claim that singular, arbitrary differences among societies that are not well-integrated parts of that society's ways of doing things are not what we mean by culture. And if one accepts that then the technique presented above can capture cultural features and measure differences among cultures' way of doing things. Now I need some data to show it.