Thanks for additionnal. Article is good and the report has very good insights.
However, need to mention some different perspectives.

Expected Cost vs Expected Utility (profit, divers, …) works when there is statistical significance on the estimations. Even though one can select to go higher cost due to operational, framework setup (ie cash vs margin ).

But it does not mean, that we should not assess the “utility” of some assets investment. But, operational setup contraints significantly the actual evaluation.

By Bias, means when N<< 100 (but one can still use Bayesian Prior).

Knowledge is not necessary derived from some actual samples (stat learning), but also for example, from logical inference,…. If A then B) …

My point is what it differencies low frequency investmsent managers vs medium/high frequency mgt is the realtive lack of statistical samples who helps deriving their knowlege.

Happy to discuss more on this topic….