Theory & Empiricism, Chicken & Egg
Growing Meta #12
Hello friends —
I’ve never been this calmly excited. I’m working on a product which is combining my love for human-centered design, and knowledge creation. The best type of product to work on, is the one you’re a potential user for. It solves a huge problem for me and I know it will help other knowledge workers.
Working on this product prompted me to rethink theory and experimentation.
We develop theories, based on logic mainly, to reach an ‘absolute truth’.
We collect empirical evidence (experimentation or observation) to understand a topic better.
Both are things that everyone does in everyday life, not just scientists. This is how we build our knowledge.
Product managers/entrepreneurs/business owners know this the most (after academics). Their works are based on assumptions (theory), or experimentations and observations. Let’s take an example on what Uber founders could be thinking when they decided to build a product: They can logically argue that “conference attendees will want an on-demand cab because it is difficult to hail for one.” Or, they can observe, or experience, how not having cabs leads to frustration.
I think this example is a bit sketchy but you get what I mean.
I wonder, what comes first, theory or empirical evidence?
What is the real relation between theory and empiricism? It has always bugged me as a chicken-egg problem. Doesn’t empirical examples lead to inferred ‘truths’/theorems, and vice versa?
It appears that philosophers have different perspectives.
So the short answer is: it’s complicated. Yes, they feed into each other, but they are independent.
In this 1988 paper, Practice, Reason, Context: The Dialogue Between Theory and Experiment [1], Lenoir gives an account of what science philosophers think about the matter. Here are some arguments on the relationship between theory and experiment:
History of science was mainly history of theory.
“Hanson and Kuhn emphasized the theory-ladenness of observation — all observation is shaped by reference to theory.”
Hacking disagrees, he notes that there are examples where “observation do not need to be inferences from theory” (e.g. “the phenomenological laws of solid state physics were known before any theory existed to coordinate them.”)
Hacking thinks that both theory and experimental practice could have their own autonomies. “Experiment has a life of its own”, someone can have an investigative goal that is driven by curiosity not to test a theory.
So, we collect evidence to understand the rules. We use rules to explain our observations & experiments.
That’s the pragmatic convention — as far as I know.
But in theory (pun intended), they are independent.
Here’s my attempt at pop literature:
Personality Test:
How do you build your knowledge: Are you a theorist, or an empirical person (or both)?
Do you like thinking what is the best way, or what works?
Wikipedia, or Wikihow?
Is your non-fiction bookshelf more about ideas or biographies/anecdotes/how-tos?
Did I nail it? Do you think HR officers in research labs will use this? Probably just missing a name like MBTI..
Calling NEPI, Najla’s Epistemic Profile Indicator (TM).
On a personal, side note:
In doing my thesis two years ago, I ended up running a lot of messy experiments to see how a model reacted to noisy data. It was extremely thought provoking. But what I really wanted to do was form a theory of representational learning: how deep learning models learn representations. My supervisor laughed.
Today, I also laugh. Developing sound theory requires solid understanding of the landscape of knowledge, and how it was formed. It is way out of scope for a masters student.
Constructing theory requires a knowledgeable diver, collecting evidence is for grounded swimmers.
[1] Lenoir, Timothy. "Practice, reason, context: The dialogue between theory and experiment." Science in Context 2.1 (1988): 3-22.