20210211 Of Bulls and Bias. : Page 2 of 7

20210211 Of Bulls and Bias.

Nigel Morris-Cotterill

Hello and welcome to The Financial Crime Risk and Compliance blogcast with me, Nigel Morris-Cotterill.

This episode is published the 11 February 2021.

It is available in text on my blog at www.countermoneylaundering.com and as a blogcast at financialcrimebroadcasting.com

In this episode:

1 Goodbye Rat. Hello Bull.

2. OFAC’s fuzzy logic

3. [un] conscious bias : it looks simple but it isn’t.

4. How to not make a fortune.

5. Corruption as explained to Sherlock Holmes.

6. The last word on rats and bulls.

2. The USA’s Office of Foreign Assets Control, a.k.a. the sanctions police and Fuzzy Logic

I remember the days when fuzzy logic was new.

It was in toasters (warm, burnt and who the hell cares so long as the butter stays on) and washing machines (whites, coloureds and how pink do you want your underwear?). It was even in rice cookers (white rice separate grains, congee or sticky rice).

In the US the Office of Foreign Asset Control, known as OFAC, which manages sanctions listing and enforcement for several agencies, has announced that it is to apply fuzzy logic to its watch lists. The results will be :

It's a hit, it's a miss, it might be here somewhere hang on while I look for something like it.

It sounds as promising as the request from an e-learning administrator in the early days of Quick To Learn More. "The passwords are too precise," he said. "Can you make them a bit, er, vague?"

The thing is - fuzzy logic is the only way to effectively search free-form data, especially that in which there is a plethora of transliterations, which is, of course, the toast and butter of sanctions lists.

Transliteration is when a word is translated from one language to another but different people spell the translation differently. Often, it’s due to an attempt to spell a name phonetically in the chosen language.

When the Arabic name Mohamed is translated into Latin characters – as used in French, German, Italian, English etc, there are nine alternative spellings. So a person who has one name in Arabic can have multiple ways in which people – not necessarily himself but it may be himself – spell his name.

The problem is not something new – indeed, if you were to look at OFAC’s lists, it often has entries that list a principal (the list preparer thinks) spelling followed by several alternatives which, somewhat insultingly, I feel, are referred to as ″also known as″ which implies that it is somehow the target’s fault that we Greco-Latin types can’t agree on a translation of his name. It means that a simple string-search simply runs over the many published alternatives. A string search is a search where the computer looks for known series of alpha-numeric characters and produces a record if and when it finds it that string.

However, because OFAC is a government department, it should, at least in theory, only publish aliases i.e. alternates for which there is evidence that it is being used and not alternative spellings which could easily lead to false positives.

That is likely to happen because, in the case of Mohamed, it’s been such a long term issue that families choose the spelling they want when naming their children. And there are many, many males and more than a few females with the name Mohamed.

Of course, this doesn’t only happen in the case of transliterations. It also happens when parents, for whatever reason, choose a traditional boy’s name for a girl, or vice versa, or who use a novel spelling for a well known name. And then there are variations on a theme, for example, Jane – JANE and Jayne – JAYNE. Add in the fact that the name JEAN is female in English but male in French and the opportunity for confusion and mis-match is extraordinary.

So, there is no doubt that OFAC is doing the right thing.

My only question as to the implementation by OFAC is this: Fuzzy Logic was first described in an academic paper in 1977. If it was in my toaster and rice cooker in the 1980s, how come you are only getting around to it now?