Predictive models: who shouldn't be using them?

By | May 9, 2012
Predictive models have been around for years. They are also rather accurate these days – 24 hour weather forecasts have become much more accurate, fish population models seem to be accurate enough to draw attention to the problem of overfishing, and I have heard rumors of Google's 'list of employees likely to leave' predictions being surprisingly accurate too. That brings us to the question: with everyone investing in crystal-ball IT these days, are there groups that should be banned from doing so?

I think the answer is yes. Using such models to determine the likelihood of a person leaving a company for HR purposes (such as who to fire next) is one example of an area in which predictive models are less than welcome. Using them as part of government internal operations against its own people is equally unwelcome: given how many models depend on neural networks and proprietary algorithms, there's a significant lack of transparence that is very easy to abuse. An 'insider-tracking predictive model', of any sort, is a worrying idea when run by a government which has lost the trust of it's people. Actually, any organization which stands for integrity and transparency, by own will or by nature of their pre-defined role, should be extremely cautious with such models, and have a carefully thought out system set up to deal with false positives before implementing predictive models and making them part of their daily business. 

The Way The Pentagon Is Predicting Your Potential To Become A National Threat Is Frightening
Guilty until proven guilty.

One thought on “Predictive models: who shouldn't be using them?

  1. Martin Watson

    If it's the government, then they shouldn't be using it. They have proved time and time again that they cannot be trusted at all with anything.


Leave a Reply

Your email address will not be published.