How Should Bail Algorithms Be Used

Should bail algorithms be used today or in the future? Here's my stance. Written for Tech Roulette 2021, P4M3 - Justice Matrix

Published: 2021-07-10 GMT; Authors: Clayton Hickey

This article is written for the third module of Tech Roulette 2021, Justice Matrix - Project 4. In the last module, we looked at how I would design a theoretical bail algorithm. In this post, we're going to look at how these bail algorithms should be used in courts, if at all.

Pros and cons of bail algorithms


  • Fast decisions
  • Consistent


  • Biased
  • Usually no better or worse than people
  • Requires lots of data for any accuracy
  • Often misused
  • Often overpromised on accuracy

My stance

Both judges and bail algorithms are biased. From the sources provided, it seems that some people seem to believe that algorithms are less biased than people are. Their only advantage is that they're consistent - even if they're consistently bad - and that they can give their answer faster and cheaper (after it's developed). I don't think that bail algorithms should be completely abandoned because they're biased. As with any technology, they are getting better at what they do but clearly, right now, they are not up to the task - at least by themselves. A judge is supposed to give a bail amount with as least bias as possible with the data they have. A bail algorithm's estimate WITH the knowledge of how the bail algorithm was designed, behaves, and works could still be useful data to a judge in making their final decision but should not be mandated in any way. For now, that is the only place I see for these algorithms. Otherwise, they should not be used until there's massive improvement.