Bias in computer algorithms - the case for and against government regulation

I'm in the process of moving my blogs to Google's Blogsite.  The following post was originally dated March 7, 2017.

--

This blog is in response to an article on the TechRepublic on February 10, 2017

TechRepublic article:
http://www.techrepublic.com/article/biases-in-algorithms-the-case-for-and-against-government-regulation

Algorithms are almost always proprietary, for a variety of reasons (e.g., competitive advantage).  The same can be extrapolated to any source code of any system, engineering designs, new product development, and more.  If the algorithm is made Open Source, then the developer needs to understand that they forfeit their claim to that technology into the future.

It is incumbent upon the developer (and his employer) to ensure that the algorithm is correct, unbiased, and does what is advertised and only what is advertised, just as it is for any software application.

Algorithms SHOULD NOT be regulated.  The only alternative for the consumer is an escrow arrangement, in which the source code is maintained in an independent third-party facility (usually to be accessed only in the extreme case of business failure, or by court order).  This is expensive, cumbersome, and a maintenance nightmare, judged by many as not viable.  Examination and evaluation of algorithms under non-disclosure agreements are a weak alternative, and not enforceable in the long run.

Unless they are specifically Open Sourced, no company is going to open their algorithms for examination or regulation.  This would be tantamount to giving away the "Keys to the Kingdom".