Joy Buolamwini: How Do Biased Algorithms Damage Marginalized Communities?

SSG Robert Ricci Excellent Professional Enlightenment!

PO1 William "Chip" Nagel thank you. When I had to work CQ or a static post I would read programming books. I would write down basic routines or functions as they're called that returned a value. That's essentially what a function versus a routine is. It returns a value. So I would build the building blocks and then when I would get off duty at sit down in front of the computer and start putting the pieces together. I have written many, many complex applications including one that was adopted by the County of Los Angeles for entry of traffic citations. I then got into writing dynamically driven websites that based upon user input the website would render itself based upon their unique keystrokes. It was in the mortgage industry where I had to pull from database tables from across the industry and come up with variable rates and comparisons for prospective clients. It was quite lucrative but it's not what I wanted to do. It's just a simple matter of a flowchart. That's a basic algorithm.

You set the parameters of the algorithm up at the beginning with either neutral values or assumed values in advance. Assumed values could be entered based on prior history of for example a Twitter user known to be a liberal. Automatically he or she is given an additional 100 points. That conservative is subtracted 50 points. Automatically there's a deficit. You then add to or subtract from those values. At the end you have a score and then you will evaluate the score based on steps. If the score falls into a particular area which may have subcategories such as hate speech you then tell the overall application to flag the post.

Quite simply it's not difficult. Artificial intelligence. All you have to do is give it its rules to begin with and let it go. Like a profanity filter. If I build a list of naughty words and I go through your post word for word and catch any of those words I can flag your post. That's where they go wrong.

The rules they have given it are slanted or weighted towards the liberal side. One can also penalize a user by doing keyword searches. If it catches a thread that mentions illegal aliens it can then subtract points. If it mentions the border wall it can subtract additional points. It reads it word for word and Loops through until there are no more words left. It renders the score. Some words you can tell it to weigh more heavily. Or phrases. You get the idea.

What may have begun as a positive post in the case of the CBD assistant director this last couple of days ended up being flagged as hate speech because the Twitter algorithm did just that. It flagged keywords and subtracted points or added points in the negative direction. The bottom line is is he had a score in a particular direction and it caused him and the New York Post to have their accounts frozen.

Have I bored you yet?

SSG Robert Ricci - Not at All Shh! Don't be Giving away the Tricks of the Trade! LOL! Do You Think, My Politics have always been static? 21 Years in Electronic and Information Warfare. My Attitude has Changed. Calling me a Liberal Evangelical would Not be Shy of What I Have Become.