Battle(?) of disciplines
Recently I was suggesting a really smart friend that she take up law. That lead to a conversation around which disciplines are ‘good’ to study and be a part of. And there is no way to answer that question without making a value judgement on the disciplines’ internal and external worth. So she naturally asked me which discipline or discipline(s) do I think is/are the most important. And if you know me, you already know the answer. I said science.
If you enjoy reading pieces like this, do consider subscribing to this newsletter to receive interesting posts like this straight into your inbox. I write about a range of issues from ideas, mental models, book reviews to policy analysis, law and technology etc. It’d be great to have you over!
Battle(?)
As soon as I said it, she jibed if I was like the average Indian uncle who thinks the only worthwhile thing to do is science and engineering. Unfortunately enough, this statement isn’t as much a stereotype as much it actually is based on truth. But we’ll get to why Indian Uncles and Aunties love science and engineering so much (while being absolutely anti science in almost all facets of their life) sometime later.
In my opinion, you must - no, you have to start from the assumption that all disciplines are inherently equal in so far as any metric to judge the value of the discipline itself will suffer from some sort of measurement myopia, exclusion, lack of rigor, value prioritization etc. Apples and oranges basically. Sure you could rank disciplines on either one or two axes, but the choice of the variable to be measured on the axes will be morally arbitrary to say the least. What I mean by that is if I try to evaluate whether Geography or History is a ‘better’ discipline, I might chose two metrics to plot their values against - say ‘employability’ and ‘contribution to GDP’. The first difficulty we encounter is definitional i.e. how do we define what we mean by our metrics - in this case employability and contribution to GDP. Honestly, the first metric is still somewhat easier to measure - you measure the number of graduates in the disciplines and then see how many of them were able to secure a job because of that degree. But as most people know, a lot of jobs that people secure are a function of a lot of other things than just the discipline of their degree. So that’s muddy waters right there. However, where we actually encounter big boy problems is in the second metric - ‘Contribution to GDP’. What is meant by contribution, what kind of, by whom, over which time period, and so forth - are questions we will have to answer to be able to make any measurement. Not only is this problem definitional, but also suffers from non objectivity due to calculation methods. I don’t even want to venture into why GDP calculation is a farce.
The aim of this exercise is to show how difficult any real value comparison between disciplines is. Not to say that it cannot be done, or that it should not be done. Just that any such exercise is going to be very difficult, and almost always incorrect.
So, does that mean there can be no meaningful discussion or judgement on the value of disciplines? There can be, but only with the caveat that I started this post with - that all disciplines are inherently equal. Because without that, all such discussions get lost into is the default ‘science is the best’ vortex. My issue with most people making that argument is that such a value judgement does not come from any rational prioritization of outcomes or optimization of variables - it is a mere regurgitation of what others are saying. No original thinking. And this is where my friend’s ‘Indian uncle’ jibe is completely true.
So then why do I think that science is the most important discipline? What better claim do I have than Indian uncles to make that argument?
You see, my claim is built on optimization and prioritization. The first thing to understand is that no value judgement of the value of disciplines can be stated as a ‘fact’. And that is because of the impossibility of comparison. So any such value judgement only rests on the kinds of values and types of outcomes that an individual prioritizes i.e. it is a personal preference and/or ranking rather than an objective judgement.1
What I mean by that is that different disciplines optimize for different kinds of things. Fields like sociology, history, literature help in better understanding of the world around us - its processes, themes etc. So if you want to understand the world you live in, study the arts and humanities. If you want to understand how the world works, how people react to incentives and make decisions under constraints - study economics. If you want to solve world’s biggest engineering and supply problems - study engineering. If you want to advance the precipice of our understanding of the physical world - study physics. The point I am trying to make is that there is no point in ranking the disciplines based on what they optimize for, because the variables they optimize for are all morally and epistemologically equal. What is important, and the only source of individual judgement on what constitutes a ‘better’ disciplines, is based on which variables you want to optimize for. Perhaps we should abandon the term ‘better’ (because it implies plotting all discipline on some sort of uniform scale) for the term ‘important’. The reason for the change in name is based on the idea that when I say Science is the best discipline, I am not arguing that Science is the best because I compared all disciplines and science is objectively better than all others in all ways, but, that the reason why I think it is the most important discipline is because I think it optimizes for the kinds of values and outcomes that I prioritize in value and outcome space.
The problem with linear optimization
The same friend happens to work with a certain government and has had practical experience of technological solutionism of deeply social problems. She told me anecdotes about how technology that appeared to accurately address the problems on paper, when deployed, was not as effective or useful as thought. And that’s the problem with linear optimization. The problem is that it is linear. You cannot have linear solutions to non linear problems, and all real world problems are non linear.
One term I like to use when I talk about optimizing for certain variables and outcomes is ‘optimization spaces’. It is this metaphysical idea space which contains the associated attributes and conditions that need to be addressed in order for certain select things to be optimized. What this means is that the optimization space for ‘technological advancement’ contains all things that need to be solved to optimize technological advancement. If it were a set, it would not be a universal set. Now the key insight that needs to be understood is that optimization spaces, even though distinct in what they optimize for, are overlapping with other optimization spaces. They are overlapping precisely because of the non linear nature of the problem statement. If it were linear and simple, we would only need to work in a linear fashion - do A, then B and then C and we would have solved the problem without any unintended consequences and unforeseen harms. But that is not how real world functions. There are hidden costs and benefits, Unseen effects and seen ignorance.
The learning that needs to be taken from the idea that optimization spaces are overlapping is that real world problems cannot be solved by linear optimization. A classic example of what I mean by this is the technological solutionism to social problems. As Ajay Shah put it, ‘Technology is a necessary but not sufficient part of the solution’. Aadhar is a perfect example of this. On paper it was the holy grail of eliminating all corruption and ensuring transparency and inclusion. But only focusing on enabling Aadhar based authentication and its associated technology without understanding the social context in which the solution was being offered was not and still is not, enough. With Aadhar, not many accounted for the fact that many of the beneficiaries of PDS are daily wage laborer’s whose fingerprints get damaged with time, thus reducing their Aadhar based authentication rates - which means that they have to forego their ration for the month in which authentication can’t take place. Many other practical problems such as not accounting for power failures, server loads, digital literacy, old age problems etc. still prove to be a hinderance only because of Aadhar based authentication - a technological solution to a social problem. Or in other words - linear optimization of a non linear problem.
Make no mistake. I am not arguing that Aadhar is bad in concept.2 But to implement Aadhar without accounting for other things that we have to optimize for is just lazy problem solving. Sometimes fatal too.
Nothing is linear. Optimize for systems and not variables.
Someone can argue that chocolate ice cream is the best ice cream. Chocolate ice cream can be compared to vanilla ice cream surely. We can graph them against the number of sales, density of consumption etc. (variables on axes) and try to decide which is the ‘better’ ice cream. But any such calculation will not be able to take all variables that can be plotted on a high enough dimensional space - due to the impossibility of complete knowledge. So in the end all you can say is that you personally think that chocolate ice cream is the best ice cream because of x y z reasons, and those x y z reasons are personal reasons and metrics - not universal.
Still waiting for a data protection bill.