Re: Law Enforcement Algorithms

Carl, while the StratML standard doesn't explicitly address hypotheses 
or assumptions, in StratML Part 2 <PerformanceIndicator>s of the 
<TargetResult>s type are analogous to those concepts -- in the sense 
that the envisioned (hypothetical) results are achieved (proven) or not 
(disproved) or perhaps only partially achieved.  They can be documented 
as <ActualResult>s and compared to the <TargetResult>s.

The degree to which objectives are achieved is analogous to the concept 
of degrees of confidence with respect to the probability that a 
hypothetical result could not have occurred by chance. However, to the 
degree that inputs and processes are documented in an open, standard, 
machine-readable format like StratML, AI can be applied to learn and 
inform users what is required to achieve the results they desire and, 
eventually, to deliver it exactly when and where it is needed (with 
close to 100% confidence for routine purposes).  When that occur, the 
advertising, marketing, and acquisition/procurement paradigms will be 
transformed.

With respect to assumptions, I like the saying about what they make of u 
and me. https://www.urbandictionary.com/define.php?term=Assume In short, 
I'd much prefer to document and share target and actual performance 
indicators than assumptions.

At least, that's my hypothesis for now.  The proof will be in the 
puddin'. https://grammarist.com/usage/proof-is-in-the-pudding/

BTW, the contents of GAO's report are now available in StratML format at 
https://stratml.us/drybridge/index.htm#AUFLE

I see that GAO plans to issue a second report with policy 
recommendations.  When they do, I plan to render them in StratML format 
as well ... but it certainly would be nice if GAO were do that 
themselves.  They do have a database to track their recommendations, at 
https://www.gao.gov/reports-testimonies/recommendations-database/ 
However, if it is interoperable with their reports, that would be news 
to me.

This report has not been categorized under the Artificial Intelligence 
<https://www.gao.gov/reports-testimonies/recommendations-database/?q=%22Artificial+intelligence%22&field=thesaurus_ss&list=1&rec_type=priority#results> 
facet of their hypertext browse index at 
https://www.gao.gov/reports-testimonies/recommendations-database/?browse=thesaurus_ss&rec_type=priority#results 
but that's understandable because it does not contain recommendations.

Owen

On 5/14/2020 3:35 PM, carl mattocks wrote:
> Thanks for sharing the report..
>
> I noted it mentioned use of Hypothesis to explain how outcome could be 
> interpreted . Would we want to note those statements as Assumptions 
> used for a Goal?.
>
> Carl
>
>
>
> It was a pleasure to clarify
>
>
> On Wed, May 13, 2020 at 11:50 PM Owen Ambur <Owen.Ambur@verizon.net 
> <mailto:Owen.Ambur@verizon.net>> wrote:
>
>     The U.S. GAO's new report on Algorithms Used in Federal Law
>     Enforcement
>     is available at https://www.gao.gov/assets/710/706849.pdf
>
>     This report is different than typical GAO reports, which usually
>     contain
>     recommendations.  So I have not converted its relevant content to
>     StratML format as I have done with the recommendations set forth
>     in 19
>     of their other reports at https://stratml.us/drybridge/index.htm#GAO
>
>     However, if anyone thinks it might be useful to do so, I'd be
>     happy to
>     give it a shot.  Goals, objectives, and stakeholders are implicit
>     in the
>     text of the narrative report.  Rendering them in StratML format would
>     make them more explicit, in machine-readable format.
>
>     Needless to say, as a self-avowed leading practice organization, it
>     would be good if GAO could start showing Executive Branch agencies
>     how
>     to comply with section 10 of GPRAMA and the OPEN Government Data Act
>     (OGDA) by publishing its reports in machine-readable format
>     conforming
>     to an internationally standardized schema.
>
>     Owen
>
>
>
>