On Patterns

Posted on

The digital humanities is fundamentally concerned with the question of how we can take what computers and discern and translate this into meaning. The main argument for using computers in this sort of research is that computers are good with lots of data and finding patterns in it. If the humanities fundamental involves, as some have argued, finding patterns then this seems to be a good fit.

But patterns themselves are just more data. We don’t know ipso facto how to interpret them. Indeed, I’m thinking of John Nash as portrayed in Beautiful Mind. You know the scene, where John Nash’s wife and colleagues see what he has been working on. They know instinctually that his work – his pattern finding – has gone from genius (he is known for important contributions to game theory, differential geometry and partial differential equations) to nonsense, paranoia, schizophrenia. If the computer similarly presents us with a pattern in huge amounts of data how do we know if our analysis is genius or crazy?

a_beautiful_mind_2

As with so many things, the answer seems to be context.

Of course the question can also be pursued on a more philosophical level. Are the patterns in the sources, that are themselves evidence of the past, or are we analyzing the past or the patterns that the computer has discerned? As Alan Liu argues, what if, through digital technology, we’re revealing patterns that are “not so much true to preexisting signals as rifs on those signals.” The computer tells us that there is a pattern in the data that we’ve fed into it, and we can choose to act on that or not. Do we accept that the patterns that we’re studying might be just that, contemporary creations?

This is necessarily complicated in the case of history. Many historians argue that they use computers to discern patterns in sources that are evidence of the past. So the patterns themselves are not necessarily interesting (nor are the claims of comprehensiveness that the statistical treatment that results in the patterns is based on). It is not the patterns that are being analyzed; they’re being used to understand the sources themselves and hence the past. So we have the argument that computers can be used as tools to facilitate the formation of knowledge according to the methods of history.
If we’re using patterns to find sources then the problem becomes more complicated still.
There are consequences for tool development of the position that we take on this issue. Because finding patterns – if we accept this as a reasonable description of what non-keyword-searching analysis is – mostly doesn’t depend on human input. Computers, unlike humans, can deal with a lot of data, so they can use it all, instead of having to filter it for relevance (although the fact that it treats all data as equivalent needs to be considered, as well as the lost of weak signals, as it were, to strong signals ). So there are two roads that development is going down: a path that depends on humans, values their abilities as a check, and one that doesn’t, that seeks to avoid their limitations. In an ideal world, the two would be integrated, the computer would compensate for human weakness, but so far this hasn’t been realized.

Sidebar