I'm very new to this field - technically not in it but want to be. I'm very early in my academic career (sophomore at a community college) but decided that I want to add a math major along with my computer science major in order to really dive into the field (when I transfer to a 4 year that is).

However - I used to be a social science major a lot time ago (before I left school and decided to come back), and some of the concepts have stuck with me from my old major. I decided to throw myself in the TCS field now and in the future, but I also believe that TCS could provide insight into some of the problems I encountered in social science.

The problems themselves aren't important to mention here, but the aspect of trying to turn social science phenomenona into isomorphic TCS problems (which could then possibly have some mathematical framework that then could formalize and clear up some of the qualitative vagueness I ran into as a social science major!) is heavily appealing - not to mention it would really challenge me as a TCS/mathematician intellectually, which I do appreciate.

Now you've read the above and have some insight, does that seem naive or stupid? I have worries I will be laughed out of TCS conferences ("look at this guy trying to turn gender into a computation problem!") for even trying this. Will I be taken seriously (I do have to mention that I will, under no circumstances, say such foolishness as "gender is Turing complete") or will I be seen as a 'waste of a PhD' ( I do plan on getting one of those in TCS).

TCS in its own right is awesome, full stop. I can easily imagine myself doing purely TCS for the rest of my life. But is trying to apply it to non-STEM (social science and even Humanities academic problems ) a fools errand?

Thank you for your time reading this.

  • 10
    $\begingroup$ It sounds like mathematical modeling/scientific computing might be a better fit for your vision than theoretical CS? CS Theory is focused on proving things about extremely precise, well-defined problems. Taking fuzzy concepts from other fields and imposing enough structure on them to make them amenable to computation is a very noble and worthwhile endeavor, but it doesn't sound to me like Theory. $\endgroup$
    – user168715
    Aug 16 '21 at 13:30
  • $\begingroup$ @user168715 I feel that this is the best answer. If you just happen to have some links at hand to useful resources for the OP to learn about what mathematical modeling and scientific computing are and how/where to learn more about those fields, I'd suggest you add those and make it into an actual answer. $\endgroup$
    – Vincent
    Aug 16 '21 at 14:28
  • $\begingroup$ I'd recommend you look into the field of multi-agent systems, which I like to think of as the intersection of computer science and behavioral economics. The stable matching problem is an example of a problem in this area. A lot of the work in this field is about designing algorithms that find incentive-compatible solutions to problems like these. $\endgroup$
    – Finn
    Aug 16 '21 at 15:07
  • $\begingroup$ Just noting that the OP has changed considerably (original title was "Is there a 'stupid' question or 'stupid' research field in theoretical computer science?"). This renders my answer somewhat moot. $\endgroup$
    – Aryeh
    Aug 17 '21 at 9:42
  • 1
    $\begingroup$ @Aryeh You are supposed to answer the question, not its title. The question didn’t change at all. $\endgroup$ Aug 17 '21 at 10:31

I won't say it is impossible, but to me it seems... challenging at best. Social science deals with the behavior of humans, and humans are complex. TCS deals with mathematics and computer algorithms, which can be precisely and rigorously modelled. Those two don't seem well-matched to me. My suspicion is that it will be difficult to apply techniques from TCS to human motivation and behavior. I'm not saying there is no hope for your vision, but I'm suggesting you inform yourself about the challenges and have a thoughtful plan for how you plan to address them.

  • $\begingroup$ Thank you for your insightful comment. Couldn't TCS be used to formalize some axioms that are noted from certain social phenomena and then jump off into forming mathematical theory from said axioms? In fact, couldn't you use use the process of reverse mathematics to determine axioms (from a quantitatively weak qualitative framework in social science/humanities) and then work from there? $\endgroup$ Aug 18 '21 at 2:07
  • 2
    $\begingroup$ Social phenomena do not admit axioms - they are inherently too subjective and vague to describe that way. Consider Richard Feynman's remark: “We can’t define anything precisely. If we attempt to, we get into that paralysis of thought that comes to philosophers… one saying to the other: you don’t know what you are talking about! The second one says: what do you mean by ‘talking’? What do you mean by ‘you’? What do you mean by ‘know’?” $\endgroup$ Aug 18 '21 at 17:08
  • $\begingroup$ You can still model people with axioms and make good predictions. Seems like this line of questioning will ultimately lead you to the field of economics, however, rather than TCS $\endgroup$ Aug 21 '21 at 20:31

There are some interdisciplinary areas in CS that may be similar to what you're looking for. These areas are often considered "applied", in the sense that they restrict their research to a certain application domain of CS and the relevance of results is mainly measured by the impact of the result on the application domain. However, these areas are also at times "theoretical", in the sense that an effective way to achieve results is applying the techniques and rigor from TCS and mathematics in general. Of course, this means that most projects will eventually have to produce a software implementation, unlike most areas in TCS.

While it is true that human behaviour cannot (at least, at the present) be fully captured by rigorous models, the increasing rate of digitization means that mathematical models will nevertheless be implicitly applied to human behaviour. Therefore, we can either study the models in order to select models that are the least bad, or hope that the models which are implicitly constructed end up being good.

One group of areas deals with computation involving models that arise from studying or managing human behaviour. Economics, as mentioned, is an example. Matching theory (with problems such as the stable marriage problem) in particular has received plenty of attention from economics, CS, and combinatorics. Somewhat related is the area of computational social choice, which deals with questions such as voting and fair division of goods (see e.g. the Handbook of Computational Social Choice (2016) )

Another group of areas is human computer interaction. Take visualization, for example. The goal of visualization is to somehow produce images that allow humans to learn from data. As such, the final test is always whether the visualization is of use to the human reading it. Nevertheless, we have enormous and complex data that we cannot visualize by hand, so we again have no choice but to make a model of the features that yield a "good" visualization. Some (often) more theoretical areas related to visualization are graph drawing, geovisualization, and computational cartography.

Visualization deals with humans interacting with the output of a computational process. The input (or entire interaction loop) the human provides also at times leads to interesting theoretical questions. The field of digital humanities has some examples, mostly in information retrieval and (possibly interactive) querying.


Yes, absolutely! This is a blossoming new field in the social sciences that a lot of people are excited about right now. The phrase to look for is "computational social science".

There is a a lot of work in graph theory and networks in particular. Some researchers to look into are:

Additionally, there is a lot of work on developing new deep learning methods to analyze sentiments in text analysis, e.g., social media posts. There is a lot of theory behind these new methods. For example, a popular one is sentiment analysis.

Another topic to look into is category theory. If you want to build purely theoretical mathematical models of social phenomena, that's probably going to be the way to do it. E.g. a paper where they did just that.

So yes, there are many reasons theoretical computer science is important in the social sciences, and we are just beginning to understand how they can inform each other. This is an exciting time to be in your position!

EDIT: Fixed the link, thanks Damiano Mazza!

  • 1
    $\begingroup$ I'm not very familiar with category theory. Could you elaborate what the advantage of the language of category theory is for social sciences over the language of graph theory? $\endgroup$ Aug 17 '21 at 8:19
  • $\begingroup$ @Discretelizard: The premise of your question is misguided; category theory is not an alternative to graph theory. In this context, its main benefit of it is that it gives a uniform language for expressing constructions from game theory, automata theory, control theory, and spectral graph theory, and furthermore this language is amenable to analysis using linear-algebraic techniques (i.e., what category theorists call monoidal categories). It's the same techniques, but expressed in a way which generalises more easily. $\endgroup$ Aug 17 '21 at 8:40
  • $\begingroup$ @NeelKrishnaswami I see. Perhaps it would have been more correct to ask what category theory can give us in this context that graph theory cannot. Nevertheless, you answered my question. $\endgroup$ Aug 17 '21 at 8:43
  • 6
    $\begingroup$ I do think that category theory has the potential to offer interesting perspectives for linking social sciences and computer science. However, judging from the abstract (I do not have access to the full paper, unfortunately) the link given as an example of the application of category theory to social sciences does not seem to contain any category theory at all (also, when I read "recent paper" I expected something more recent than 2007, but that's subjective :-) ). Maybe there's a mistake in the link? $\endgroup$ Aug 17 '21 at 9:45

It seems to me that your best chance at applying TCS to social sciences is to apply it to study problems in economics. In particular, one can use TCS to argue that certain tasks required in economics are computationally infeasible. For a concrete example of this, see the following write-up: https://m-cacm.acm.org/magazines/2011/5/107705-computational-complexity-and-information-asymmetry-in-financial-products/fulltext#R5

  • 1
    $\begingroup$ Thanks for the advice, I appreciate it. Know the funny part? I'm actually looking at other fields like sociology/anthropology. $\endgroup$ Aug 16 '21 at 7:18

"Stupid" is subjective and judgmental, but there are certainly meaningless, trivial, or uninteresting questions. Or at least sometimes they can seem meaningless until someone comes along and imbues them with meaning. Let's take math, for example. The search for the smallest positive real might have been meaningless before nonstandard analysis was discovered. Similarly, the search for an integer that's both even and odd might seem silly at first, but perhaps might actually make sense in some paraconsistent logic.

"Trivial" means easily solvable by anyone with a minimal set of commonly available tools, not a research question. "Uninteresting" means lacking a compelling motivation or structure. Number theory is famous for being able to generate endless problems not solvable by current techniques, and number theorists are rather picky about which ones they choose to devote time to. Their criteria, I imagine, include: which are the likeliest to be amenable to the tools we do have? Which hold the promise of shining light on new structure or leading to new tools and techniques?

Finally, I'll attempt an answer to the OP. A "stupid" question is one that the research community agrees, by broad consensus, is not a research question -- and is posed, as a research question, by someone who is in a position to know better.


I used to study economics, and the answer is yes. In 2019/2020 Nobel Laureate Paul Milgrom gave the marshall lectures in cambridge with the title "Market Design When Resource Allocation is NP-Hard". I believe that there is plenty of intersection between game theory and TCS, (e.g., search algorithmic game theory books on amazon).



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.