The problem with Elon Musk’s ‘first principles thinking’

A while ago I wrote an Untangled piece called The Artificial Gaze, where I asked what society would be ‘optimized’ for if it was one big algorithmic system. Unrealistic beauty standards definitely seem to be one thing — and also the pursuit of efficiency and scale. Today I want to add another thing to the optimization pile: Objectivity and the presumption of value-neutral knowledge. Let’s dig in.

Elon Musk is famous for his ‘first principles’ approach to management and innovation. Just to be clear, he did not come up with this approach, he just champions it a lot. Musk describes the approach as such:

“Boil things down to the most fundamental truths and say ‘OK, what are we sure is true, or as sure as possible is true?’ And then reason up from there.”

This so-called first principles thinking is intended to push back on reasoning by analogy or by looking to the past, and instead, begin from the beginning. It is an attempt to unlearn and remove prior assumptions — to get those out of the way so that one can begin anew, with a blank sheet of paper. Children instinctively get first principles thinking: when a child asks ‘why’ over and over again, they’re cutting through the analogies and shared beliefs that approximate such an approach. Now, I don’t have a problem with the pursuit of unlearning assumptions, biases, and values — that is very much part of how I see the project of Untangled. But the belief that we can achieve a kind of objectivity or arrive at ‘fundamental truths’ is a big problem.

Why? Well, absolute objectivity and value-neutral knowledge have always been a myth. Sergio Sismondo, a scholar of science and technology, explains absolute objectivity as, “the ideal of perfect knowledge of some object, knowledge that is true regardless of perspective.” This notion that we can employ a detached perspective to derive independent ideas is very similar to Thomas Nagel’s conception of ‘the view from nowhere’. This presumes that the truth is out there and that it’s the job of scientists and technologists to stand apart from the world to find it. ‘First principles thinking’ and ‘objectivity’ attempt to separate knowledge from where it comes from. But that’s where the problem lies: As Professor Donna Haraway once wrote, “All knowledge is knowledge from somewhere.”

Feminist scholars have long argued that knowledge is situated and partial: that we can’t separate knowledge from those participating in its creation, and the process they pursue. Because those doing the science get to decide on a bunch of things such as what counts as a relevant research question, what counts as legitimate evidence, how to structure the research design process, how to interpret the data etc. This means that in societies structured by inequality, bodies of knowledge often represent the interests, biases, and assumptions of the dominant groups. Haraway put it this way:

“Rational knowledge does not pretend to disengagement: to be from everywhere and so nowhere, to be free from interpretation, from being represented, to by fully self-contained or fully formalizable. Rational knowledge is a process of ongoing critical interpretation among "fields" of interpreters and decoders. Rational knowledge is power-sensitive conversation. Decoding and transcoding plus translation and criticism; all are necessary. So science becomes the paradigmatic model, not of closure, but of that which is contestable and contested.”

So what would it look like if we started instead from the presumption that knowledge is situated and partial? Sandra Harding argues for “strong objectivity” in Whose Science? Whose Knowledge? which explicitly includes the researcher’s social standpoint in the scientific process. It’s not assumed away or treated like a bug, it’s a feature. Patricia Hill Collins, in Black Feminist Thought, argues that “partiality and not universality is the condition of being heard.” Right, if we assume that knowledge always comes from somewhere, then the process of science doesn’t output ‘the Truth,’ it outputs partial perspectives that should be challenged and contested. Sandra Harding summarized this idea nicely when she said “What science becomes in any historical era depends on what we make of it.”

If you believe that the scientific method produces unambiguous truths, this all might sound a little radical or unsettling. But, there are two key problems to assuming that science outputs ‘The Truth’. Firstly, the scientific method often generates more questions than answers. As scholar Nathan Jurgenson put it, “The scientific method, based in deduction and falsifiability, is better at proliferating questions than it is at answering them.”

Secondly, just by examining the process by which science is made, it’s pretty obvious that it does not produce absolute objective truths. As scholars of science and technology studies have shown, the laboratory, much like a startup or engineering team, contains assumptions, beliefs, and social dynamics — and these all shape experiments. Furthermore, the results of experiments are interpreted in multiple ways — what scholars call “interpretive flexibility” — and social groups organize around these interpretations and fight to stabilize their version of the story as ‘The Truth’. So, whatever becomes ‘The Truth’ is often more about which narrative prevails than it is about the scientific method.

At the moment it seems as though these features of science are not taken into account, and so science is often viewed as upstream and apart all things social and cultural — which is why we glorify value-neutral knowledge and first principles thinking. But if the scientifically produced ‘Truth’ is actually a matter of perspective, then we need to shift our thinking away from the idea that knowledge is objective and stable, to thinking of it as something that’s situated, partial, and socially constructed.

In order to do that, we need to relax our assumptions about the following:

• Expertise and ‘the hard sciences’: Sure, let’s save a seat at the table for engineers and physicists, but if we don’t make room for sociologists, anthropologists, political economists, and especially those affected by the sociotechnical systems, then we’re missing an important part of the table!

• That facts are stable: in actuality, they are contingent. Sure, some facts endure unperturbed but others evolve over time or change depending on the context.

• What is knowable: thus embedding more uncertainty and humility into our decision-making and policymaking. Historically, policymaking entails smart people making decisions based on a lot of research and evidence, and finding that the problem persists or gets worse. This demonstrates a need for more uncertainty and humility, but also more flexible frameworks.

Relaxing our assumptions about what’s stable and what’s knowable can be scary. It requires a letting go of control and an embracing of uncertainty. But this mindset shift can also be empowering. As anthropologist David Graeber put it, “The ultimate, hidden truth of the world is that it is something that we make, and could just as easily make differently.” It’s actually in relaxing our assumptions about what’s stable and knowable that we confront our own agency; our ability to change existing systems and take the brave steps to build new ones.