Digest: Resisting Reality and the Art of Shifting Assumptions
Coverage of lessw-blog
A look at how re-evaluating foundational premises can help engineers and decision-makers bypass 'hard' constraints without denying logic.
In a recent post titled Resisting Reality, a contributor on LessWrong examines the cognitive and practical strategies required when logical evidence points toward outcomes we wish to avoid. The piece explores the tension between accepting rigorous mathematical conclusions and the human desire to shape a different outcome.
The Context
In fields ranging from high-frequency trading to distributed systems architecture, professionals are often bound by "hard" constraints-mathematical theorems or market realities that dictate what is possible. A common response to these constraints is a form of passive acceptance: optimizing within the box. For example, in distributed systems, the CAP theorem posits that a data store can only provide two of the following three guarantees: Consistency, Availability, and Partition Tolerance. Standard engineering practice involves accepting this trade-off and choosing the least damaging compromise.
The Gist
The author argues against simply accepting a "doomed" reality dictated by current evidence. However, the proposed alternative is not to deny the theorem or the facts-which would be irrational-but to "resist the mathematics" by attacking the underlying assumptions. The post suggests that when a derived conclusion is unacceptable, the solution is to shift the foundational premises.
Using the CAP theorem as a case study, the author discusses how one might couple distributed systems (using mechanisms like sidecars) to effectively bypass the constraint. By changing the system boundaries or the definition of the problem, the engineer can render the theorem irrelevant to the specific desired outcome. This approach is contrasted with the mindset often found in finance, where professionals prioritize accurate forecasting and arbitrage over engaging with the political or structural realities driving those forecasts. The post posits that while accepting reality is necessary for accuracy, "resisting" it by altering the inputs is necessary for agency and innovation.
Why It Matters
This perspective is particularly relevant for those working in AI alignment and complex system design. It offers a framework for addressing challenges where the "optimal" technical solution conflicts with broader goals. It encourages a proactive approach to re-evaluating premises rather than resigning oneself to mathematically sound but undesirable futures.
Read the full post on LessWrong
Key Takeaways
- Reframing Constraints: When faced with mathematical impossibilities (like the CAP theorem), the goal should be to alter the system design to render the theorem irrelevant rather than accepting the trade-off.
- Homeostasis vs. Truth: The post highlights the psychological friction between maintaining a stable worldview and updating beliefs based on negative evidence.
- Agency in Engineering: Accepting "market-optimal" or "inevitable" outcomes without challenging underlying assumptions can lead to a loss of creative agency.
- Resisting Mathematics: The strategy involves respecting the logic of a theorem while rigorously changing the axioms or assumptions that feed into it.