So I need different names for the thingies that determine my predictions and the thingy that determines my experimental results. I call the former thingies "belief," and the latter thingy "reality." —Eliezer Yudkowsky
Feeding on Reality
Feedback is something that happens after we do something else first. It can be quick feedback, like your phone gently vibrating at a button push; or slow, like your awkward annual performance reviews.
This feedback gives us a sense of causation: doing X affects Y. While working, you'll observe X and Y and want to understand their relationship. A hypothesis is a prediction about how we believe this causal relationship works. The feedback from testing and refining hypotheses helps close the gap between belief and reality—a desirable position to be in when making decisions.
The relationship between X and Y can be unclear and require additional tools like Thinking Clearer With Directed Acyclic Graphs (DAGs). These help us avoid circular reasoning and distinguish direct from mediated effects. Moreover, hypotheses can be deterministic, or binary: this is true; that is false. Or, more likely, they will be probabilistic: we believe that X causes Y, and we assign a Z% probability that this hypothesis maps to reality.
Sometimes, our beliefs are sown by catchy heuristics, circulating anecdotes, or outdated assumptions, but these should be the seeds of our hypotheses, not the harvest.
Structure your hypotheses to make specific, testable predictions, which mitigates biases and allows data to persuade. Sometimes, data is unavailable due to cost, timing, or ethics. However, you can still gain insight by pondering the hypothesis's implications. A hypothesis should constrain expectation—it precludes some outcomes.
- If this hypothesis were true, how would the data look?
- If this hypothesis were false, how would the data look?
These questions play into Encoding Intuitions, and this Bayesian mindset lets you compare hypotheses and update your beliefs as you see new data. The above questions are the likelihood from Bayes' theorem: What is the probability of this data given that the hypothesis is true, false, or our prior as a statistical distribution?
Beliefs, Identities, Hypotheses
In the workplace, a hypothesis is also an invitation for collaboration. By saying, "My hypotheses are X, Y, Z," you create a safer environment for debating ideas—a scientific sandbox. You're not saying, "I believe this," but rather, "Let's explore what if this were true." Otherwise, exploring causation will feel like defending predetermined conclusions or sharp critiques of your identity. Painful.
So, be alert in considering alternatives, even if they initially seem laughable, unreal, or challenge your current understanding. Life is messy, and systems are complex; a well-supported or favored hypothesis may sour when reality is observed elsewhere.
With practice, you can start thinking in terms of hypotheses. It's a feedback cycle of hypothesizing, testing, and updating our beliefs. Moreover, it will lead to exploring new ideas, sidestepping some biases, and moving toward better decision making. A worthwhile harvest indeed.
Formal Learning Theory. In E. N. Zalta & U. Nodelman (Eds.), The Stanford Encyclopedia of Philosophy (Spring 2022 Edition). The Metaphysics Research Lab, Stanford University. https://plato.stanford.edu/archives/spr2022/entries/learning-formal/
Galef, J. (2021). The Scout Mindset: Why Some People See Things Clearly and Others Don't. Portfolio.
McElreath, R. (2020). Statistical Rethinking: A Bayesian Course with Examples in R and Stan (2nd edition). Chapman and Hall/CRC. https://xcelab.net/rm/
Yudkowsky, E. (2007, November 6). Leave a Line of Retreat. LessWrong. https://www.lesswrong.com/posts/3XgYbghWruBMrPTAL/leave-a-line-of-retreat
Yudkowsky, E. (2008, January 1). The Simple Truth. LessWrong. https://www.lesswrong.com/posts/X3HpE8tMXz4m4w6Rz/the-simple-truth
This website reflects the author's personal exploration of ideas and methods. The views expressed are solely their own and may not represent the policies or practices of any affiliated organizations, employers, or clients. Different perspectives, goals, or constraints within teams or organizations can lead to varying appropriate methods. The information provided is for general informational purposes only and should not be construed as legal, actuarial, or professional advice.
