I quickly realizes that trying to figure out the best algorithm or the best equations was not a smart way to proceed. So, instead, this model relies on me telling Gemini what I thought was important, and what I did not:

This is a procedure for finding the best match between the most recent OFAC enforcement action and all the enforcement actions in a given year.

Compare the Enforcement action document from the most recent settlement or imposition of Civil Monetary Penalty (i.e. ignore Findings of Violation) to each  enforcement action document in the time frame specified, other than the most recent enforcement action (e.g. do not compare the most recent enforcement action to itself). Omit any enforcement action which does not appear to refer to the enforcement process from the OFAC Enforcement Guidelines, and omit any Findings of Violation.

General Factors should be considered in tiers, in descending order of influence on the match result. In each tier, all General Factors should be considered to have approximately the same importance and weight in the resulting model. 

Tier 1 (highest weight/importance): General Factors A, B, and C

Tier 2: General Factors D, E

Tier 3: General Factors F, G

Tier 3 (lowest importance/weight): General Factors H, I , J, K

determine which is the closest comparison, based on the following factors:

Factor 1: A comparison of the generalized behaviors (e.g. ignoring warnings from other parties, both internal and other non-regulatory parties, like banks who reject payments for sanctions reasons; wire stripping; falsification of documents or changing information to something misleading to third parties) noted in the enforcement release. Each matching behavior increases the result of this comparison, while each behavior that only one action represents decreases its value. Additionally, each factor must be weighted based on the General Factor the behavior is most associated with. So, matching on having no compliance program means a lot less than matching on Willful Behavior.

Additionally, the ratio of General Factors to non-matching Factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

Factor 2: A comparison of the General Factors noted in the enforcement release relating to the aggravating factors noted. Each matching General Factor increases the result of this comparison, while each General Factor  that only one action represents decreases its value. 

Additionally, the ratio of General Factor-weighted matching behaviors to non-matching factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

All comparisons and calculations should take into consideration the tier of each General Factor

Factor 3: compare where each action falls on the OFAC Penalty Grid:

If the quadrant is the same, the score is the highest.

ELSE If both are “egregious”, the score is less than the previous..

ELSE If both are “not voluntarily self-reported”, the score is less than the previous.

ELSE If both are “non-egregious”, the score is less than the previous..

ELSE If both are “voluntarily self-reported”, the score is less than the previous.

ELSE the score is the lowest

Factor 4: was there a lack of cooperation, other than lack of voluntary self-disclosure, in both enforcement actions, or was there substantial cooperation, in both enforcement actions?

Weight Factor 4 at 70% of Factor 1

Factor 5: Compare the periods of time of the violating behavior in both enforcement actions 

Period Length Tier 1: longer than 3 years

Period Length Tier 2: between 1 and 3 years

Period Length Tier 3: shorter than 1 year

If both actions are in the same tier, score this the highest

ELSE If the actions are in adjacent tiers, give this an intermediate score

ELSE score this the lowest

Factor 6: A comparison of the General Factors noted in the enforcement release relating to the mitigating factors noted. Each matching General Factor increases the result of this comparison, while each General Factor  that only one action represents decreases its value. 

Additionally, the ratio of General Factor-weighted matching behaviors to non-matching factors should adjust the result – so that a 60% ratio should produce a higher overall result, all other factors being equal, than a 30% ratio.

All comparisons and calculations should take into consideration the tier of each General Factor

Take no action on this information until prompted further.

Notice that while I gave relative scoring, I did very little, if any, hard-coding.

Next: my results