Legitimizing, Yet Again, Why Digital and Computational Methods Are Valid

Like, I assume, many other people working with digital or computational methods in their respective sub-disciplines, I have recently found myself explaining, yet again, why historians might want to use computational methods and why they have a legitimate place in our field.

As I have mentioned in the post where I discussed the formation of DH subdisciplines, I recently was involved in editing a special issue on computational approaches to the history of alchemy and chemistry. In our introduction, we went through many of the questions that tend to come up in this context: What are computational methods? How can they benefit historical research? Why should historians use them? What are their limitations? What are the typical concerns and why doesn’t their mere existence render all computational research obsolete? One of my co-editors even wrote a two-part article specifically on the opportunities and challenges of computational approaches within the history of chemistry.

And to a certain degree, I think doing all this explaining is really important.

At the same time, I have to admit: I’m starting to get a little tired of having to explain why we do this at all, over and over again. I’m starting to feel like a broken record.

We do still need to have these conversations?

As I’ve discussed my earlier blog post on discipline building in DH, I don’t think a strong digital and computational sub-discipline has really formed yet within the histories of science, knowledge, and technology (my DH subfield). And that means some of this groundwork simply hasn’t been done yet. We need “normal” or traditional historians of science and knowledge to have better literacy around digital and computational methods. Otherwise, those of us doing digital and computational work don’t really have interlocutors with whom we can properly discuss what we’re doing.

I often encounter historians who feel strongly that Digital Humanities methods aren’t for them. And that’s completely fine: not everybody needs to use these methods themselves. But at the same time, I sometimes suspect that the people dismissing them don’t actually know all that much about them. Everybody would benefit from being a little more open to these discussions.

Because there isn’t a particularly strong digital and computational sub-discipline in the histories of science, knowledge, and technology, people also haven’t had that strong an incentive to keep up with these developments. Nobody is asking everyone to understand every technical detail. But compared to disciplines where digital methods are more prominent, I do feel that many historians of science, knowledge, and technology who aren’t themselves using digital methods or involved in digital projects remain quite distant from them.

Being critical of technology isn’t quite the same as understanding computational methods

The history of science (also related fields like Science and Technology studies) has, of course, a long history of criticising technology and a lot of expertise in doing so. Consequently, we now also have a strong movement of people who are very critical of AI. I think that’s fair. I have plenty of qualms with AI myself. But AI and computational methods are a complex field, and while I trust that many of my colleagues have a solid intellectual background from which to criticise AI, I sometimes feel that what’s missing is more familiarity with computational methods themselves.

So, yes, I think it’s important to explain again what computational methods are (because it’s quite simply still necessary), why some historians might want to use them, what the typical criticisms are, and what the differences are between “digital,” “computational,” and “AI.” I’m glad if our introduction can introduce some readers to these concepts who haven’t been following these discussions as closely. But it does get a little tiring. Why do I have to explain myself all the time when other people don’t?

We tried not to let that frustration drag us down. Instead, we thought: fine, we’re explaining these things anyway, so let’s actually make the explanation useful. From a DH perspective, many of the claims we’re making aren’t necessarily new. We’re spelling them out for readers who may not have followed these debates over the last couple of decades.

But we also tried to address criticisms and concerns that we have actually encountered within our own field (“discipline of origin”, i.e. history of science, knowledge and technology). One obvious example is the fear that computational methods decontextualise historical sources. That’s a relevant concern. So rather than simply dismissing it as a misunderstanding of computational research, we looked at how we can respond to that scepticism in practice. How can we design our research so that these problems don’t negatively affect the results? And where a limitation can’t be avoided, how can we acknowledge it and still work out what we can reasonably learn from the research?

It was also useful to situate what we’re doing within the existing field. You often know yourself where your work comes from intellectually, but spelling that out can help other people understand the background and discover previous work they may not have been following.

New possibilities also mean new responsibilities

In explaining the pros and cons, up and downs of digital and computational methods, we started with the very familiar claim that computational methods open up new horizons and new possibilities for historical research. That’s one of the standard arguments DHers use when explaining why computational methods are useful. They allow us to ask questions differently, work with material at different scales, and do things that wouldn’t otherwise be possible. But we didn’t want to stop at this locus communis, so we also argued that computational methods bring new responsibilities. As regular readers probably know, I care a lot about data ethics and related questions. I’ve written here before about things like data feminism, and I’m increasingly thinking about these issues as part of a broader paradigm that I would like to advance: critical computational humanities.

Towards critical computational humanities

Critical Digital Humanities already exists as a sub-field, although there are, of course, discussions about what its role should be and how it relates to DH more broadly. What I would like to see is more of that critical theory and critical reflection brought explicitly into computational humanities as well.

Digital and computational approaches are necessarily intertwined. As I’ve argued in the epilogue to our special issue (see also the blog post reflection about it), the computational humanities cannot exist without a DH basis, meaning the data work that comes before it. (I have also blogged about Why Looking More Closely at Our Data Is the Way to Better Research Ethics).

So the question isn’t whether computational humanities needs to engage with these issues. It already depends on them. And this is actually somewhere where I think the history of science can contribute something important to DH and computational humanities more broadly. Historians of science have a lot of experience critically examining instruments, knowledge production, and the conditions under which knowledge is created. Why shouldn’t we bring that expertise into computational humanities?

What kind of history gets turned into data?

In our introduction, we pointed out how the history of science, knowledge and technology increasingly understands (or at least wants to understand) itself as a field that is much more global and diverse than it used to be.

I think we need to carry that development into computational humanities as well.

If we want computational humanities, or specifically critical computational humanities, then we need much more diversity in the languages, time periods, source materials, types of knowledge, and kinds of history represented in our computational research.

At the moment, that representation is far from equal.

The data available to us are skewed towards particular kinds of historical actors and sources (for example, the people we have traditionally constructed as Western scientific “geniuses.”) Those are the materials that have often been preserved, catalogued, digitised, and made easily available. Unsurprisingly, they then remain central to many digital projects. And that skew becomes even more significant once we move from digital to computational research.

Computational research inherits the biases of digitisation

Computational historical research usually depends on pre-existing datasets in one way or another. That makes us particularly vulnerable to the biases already present in what has been digitised, how it has been digitised, and what has been turned into structured data in the first place. If certain histories aren’t represented in the data, computational methods don’t magically bring them back.

In our aforementioned introduction, we drew on Sandra Harding’s concept of strong objectivity. Our argument was that if we’re going to use computational methods more extensively, then we also take on a responsibility to make the histories we produce more diverse. That doesn’t happen automatically. It requires active steps and broadening our horizons to make our research more diverse: for instance, to include understudied groups or cultures.

How do you do that if that’s not materials in which you have been trained? I also wondered that when my boss asked me to think beyond studying the European materials that form the core of sources that I study. I simply didn’t study, say, Chinese as my degree. I studied European languages and cultures.

So how can one make one’s research more diverse?
Through collaborations, my friend. That’s how you do it. It requires some humility because you won’t be on your turf. But do it, it’s worth it. I have been trying my best and it has expanded my horizons in beautiful ways I simply wasn’t open to before. Was it easy? No. Did I get particularly far? No. But every inch matters, I guess.

So perhaps the question shouldn’t only be: Why should historians use computational methods?

We should also be asking: If we get the increased scholarly prestige from using computational methods, how can we give back?

Explaining ourselves, yet again, but maybe with a new twist…

Coming back to the issue of explaining ourselves, I ultimately think we still need the introductory conversations. In the history of science, knowledge, and technology, there is clearly still work to do when it comes to building literacy around digital and computational methods. If our special issue helps with that, I’m glad. Maybe the same is true for your subfield.

Actually, I think our introduction became a pretty good generally usable introduction into digital or computational methods for historians, if I may say so.

But I also hope we can increasingly move beyond having to justify the basic existence of computational approaches over and over again. There are more interesting questions waiting for us. Like…
What kinds of histories are our datasets allowing us to write? What has been excluded before we even begin our analysis? Whose languages and sources are available computationally? How do the infrastructures we rely on shape the research questions we’re able to ask? And thus, what responsibilities do we take on when we choose to work computationally?

These are questions I’ll certainly return to. I’m thinking and writing more and more about data ethics, critical computational humanities, and such topics, and I’ll try to take you along when I can on this blog.

Regular readers will know that I’ve struggled a bit over the last few years to keep posting here consistently. So perhaps this is also a way back into it: when I’m working through these questions in articles and academic publications anyway, I’ll try to share some lighter introductions to those ideas here or pull a few key thoughts out of the articles (like I have done here) and link to the full texts for anyone who wants to go deeper.

Maybe this gives you something to think about in your own field.

Until next time, thanks for all the fish!
The Ninja