BAE3. update or updated phi (environment) i.e. pivot

[hyunjimoon]

Discussed in Data4DM/BayesSD#192

^{Originally posted by hyunjimoon February 22, 2024}
Q1. there's difference between voluntary environment change (update phi) vs adaptation to the changed environment (updated phi). I'm curious how these differences affect the following actions and whether I can call the former expatative.

Q2.
There's no theoretically grounded definition of pivot yet and its perception differs:

• keeping the vision, changing strategy - Ries (lean startup) and Eisenmann (Harvard prof.)
• Hierarchy of startup decision: vision-strategy-roadmap. pivot is change in the vision level” - Daniel Lee (previous Generable CTO)
• framing startup pivot as p(y|theta, phi=new)>p(y|theta, phi=current), lane changing model from #186 can be relevant

context: Charlie and I are interested in defining pivot. I wonder whether the joint inference of goal and belief mechanism detailed in grounding language about belief in bayes theory of mind appear in nature. "Baldwin effect" is the closest which I'm building up below.

how humans infer other agents' goals, beliefs, and plans from their actions by modeling these inferences probabilistically. This model incorporates the concept that our understanding of others' beliefs is deeply tied to the goals we infer they have and the plans they make to achieve these goals.

Q3. has there been attempts to determine the optimal level of "empty" space (where spandrels can grow) beneficial for adaptation? e.g. can it be assessed similarly to the secretary problem's 37% rule, plus how would the environment's "hardwareness" (defined as by variance ratio of demand and supply function i.e. uncertainty in data generating process of each) affect this optimal ratio?

Q4. can we draw parallels between academic theories and physical or biological entities? i.e. hypotheses, unit theories, and programmatic theories might correspond to phenotypes, alleles, and genotypes, respectively, based on how they represent and integrate environmental understanding and observable traits.

detail below for each

[hyunjimoon]

JB, Angie agenda (apr.15)

• information organization & management (board vs discussion)

• pivot simulation (6 hypotheses e.g. 🪵 BayesSD meeting Log BayesSD#23 (comment))

• biological / evolutionary connection with pivot

• (TBC) literature review on pivot including Camuffo24.pdf which explains how education (scientific hypo.testing) affects pivots and performed RCT experiment. "methodic doubt (greater caution) and efficient search (better information). We show that fewer pivots imply that, in our sample, the former mechanism dominates." is the key

• JB's agenda