What is it and why is it important?

MY guiding phisolophy has always been, as far as is humanly possible, to ensure clarity of thought so to enable confidence the decisions which are a product of that thought and a confidence that action arising out of those decisions are highly probably to lead to a net advantage.

With the development of local-running LLMs, it is of interest to analyse decision-related data to study not just the decisions and their outcomes but events, the environment and the context of things preceding such decisions.

As it is the diary system is unstructured and does not lend itself to such analysis.

The proposal is that I develop the a decision log based off Farnham Street’s decision log template and extend their idea into something that incorporates my knowledge, skills and experience.

This is ultimately to capture the essence of conditions/systems out of which good decisions (or near-good decisions) arise so that small incremental improvements can be made.

What has been my experience with it?

While I have used decision tools ie. tools/techniques which help break down a decision problem into components and provide a way to reason about it, I have not specifically looked evaluating the quality of the decision. I have simply assumed that the more sophisticated or elegant the technique used to make the decision, it must follow that the decision AND its implementation must be sound.

There are several kinds of decisions which defeat the purpose of good making-technique, at the fore is decisions which have medium to unknown completion points. It is quite easy to lose the focus of why something is being done, or to modify the decision and plant to fit available implementation energy.

How have I operationalized this knowledge?

This falls in the area of system inputs and processes. I am attempting to ensure that what is fed into the action system is high quality intentions.

The wildest possibllities this knowledge might enable

At the extreme end an AI-agent could extract morals and values from the data and guide future decision-making, or I could build such a model(s) myself.

At the median it might unlock analysis of how the system is performing, shining a light on an unrelated performance area. The experience gained in the development of such a software system would also be valuable.