The Algorithmic Odds of Jury Duty - perhaps not so random!



How random is jury duty selection? Maybe not as random as we think.


Substantively, the research reveals that the jury selection process most fundamentally has a systemic bias which favours financially stable urban homeowners.Canadian Institute for the Administration of Justice

That might explain why I’ve been summoned for jury duty, not ONCE, but TWICE in less than six months! This will be the THIRD time in my life that I’ve been called to appear for jury selection. While I’ve never actually served on a jury, this most recent incident of being summoned twice in a very short period of time made me curious about how jury selection works and what biases - technological and social - are built into the process. I went searching for answers and came across this post on Quora:


Bias in the database of possible jurors


The jury selection process is set out in each province’s Jury Act. Here in Alberta, the eligible pool for jurors is fairly broad. You must be 18 years of age, a Canadian citizen and a resident of Alberta. There are exclusions made for certain professions specified in the act such as lawyers and elected officials. You’re also excluded if you have a criminal record for which you’ve not been pardoned.


In order to summon people who fit the approved criteria, certain lists are used to connect with this population. The Alberta Jury Act mentions two: voter registration lists (election rolls) and property tax assessment lists (assessment rolls). Right away, we see a few societal biases at this stage:

  • The qualification to be a Canadian Citizen means many minority groups can be excluded. For example landed immigrants who’ve been given the right to permanently reside in Canada don’t qualify for jury duty. This can result in less representative juries when actual selection takes place. At the individual level, you might see this as a benefit but at a societal level, it could be seen as a loss of greater diversity. Conversely, you might feel excluded from this important process even if you’ve lived in Canada a very long time.

  • If you’re on the registered voter’s list you’re a more likely candidate to be selected. In some places this caused a disincentive to be on the voter’s list, which resulted in the expansion to use other lists such as driver’s license or health records.

  • If you’re a property owner, you’re on another list which makes it easier to connect with you. This introduces a level of socio-economic bias into the process.


Beyond the above issues at the possible universe of potential jurors level, there is further bias that can arise from exemptions or requests to not take part. For example:

  • Age: You can be exempted if you’re over 65. As the population ages, this could put more of a burden on the remaining 18-64 year olds.

  • Urban vs Rural: In order to avoid hardship for those who live very far from where proceedings would take place people living in rural areas might be granted an exemption.This means over indexing on urbanites.

  • Language barriers: Those who are not able to understand the language in which the trial will be conducted can request an exemption.


Again, if you are seeking to be exempted from jury duty, these factors might work in your favour. However, at the collective level, they skew the pool of possible jurors in the direction of younger and middle-aged, urban, English speaking, law abiding citizens who also vote and own property. In other words - people like me!


In the last Edmonton election we had 37.6% voter turnout with 236,488 people voting. That means 628,957 people were eligible to vote based on the voter registration list. That list would need to be further refined to filter for exclusions for jury duty, which might be hard and time consuming to do. Let’s ballpark this and call it a one in a half million chance of being selected randomly for jury duty from this list.


Are random selection algorithms really random?


Computers are designed to carry out specific instructions. By definition, they are not random. In order to use a computer to generate a random selection, there are algorithms called pseudo-random number generators. These algorithms all need a seed value - a starting point - from which they will then work their numerical magic. How to ensure that the seed itself is random (good)? That is the issue. In non-tech terms, if you’ve got a bad seed, you’ll have a bad (not so random) outcome. Better ways of determining a seed lead to better outcomes. These algorithms approximate randomness, but in essence, there is always a sequence or pattern taking place. Mathematicians, statisticians and others in the know have warned about this issue:


The list of widely used generators that should be discarded is much longer [than the list of good generators]. Do not trust blindly the software vendors.” International Encyclopedia of Statistical Science

So, what specific software or algorithm(s) are the courts in Alberta using to generate this random jury summons process? How reliable or trustworthy are those systems? I don’t know. Neither did the person I spoke with at the Jury Management Office (though they were super nice and they did offer to take my details to see if someone else can help me).


The ACLU recently investigated certain automated decision systems used in the jury selection process in Washington state. They referenced many different kinds of system failures that disproportionately excluded Black and minority populations from juries. Excluding certain groups results in over sampling others.


The fact that my name has come up twice in a short period of time isn’t, mathematically speaking, evidence of non-randomness. If things are truly random “999” may come up as often as “721” - with randomness you just never know. But, non-mathematically? It doesn’t feel random, particularly when your name goes into the jury process more often because of the biases in the source data.


Greater Algorithmic Transparency


My personal jury summons saga highlights the many elements I focus on in my work such as biases in data that reinforce and amplify problematic societal patterns and our blind trust of computer systems and algorithms - even when experts acknowledge there are problems. Most importantly, there is no transparency about these algorithms that are being used to make important decisions for individuals.


Showing up for my next scheduled jury selection date is a small inconvenience (privilege?) for me. There are other algorithms used in the justice system - like the infamous COMPAS software - that have much greater consequences for people. We need transparency to know when algorithms are being used. We need some level of explainability to understand how the specific system works. We also need to be able to contest the decisions made by algorithms if we believe they are flawed.


A sad footnote - I wanted to connect with Dr. Lane Mandlis, the University of Alberta professor who answered the Quora question. Sadly, he passed away. His story highlighted a number of inequities for disabled persons and end of life care in the province of Alberta. It’s definitely worth a read.


Resources


For history lovers, a John Von Neuman paper in which he states “there is no such thing as a random number” and the math to back it up.


Here’s a blog post that goes into more technical details about how these Pseudo-Random Number Generators work


The whole series by the ACLU How do automated systems affect our lives? is excellent! Well worth a read.


By Katrina Ingram, CEO, Ethically Aligned AI


________


Sign up for our newsletter to have new blog posts and other updates delivered to you each month!


Ethically Aligned AI is a social enterprise aimed at helping organizations make better choices about designing and deploying technology. Find out more at ethicallyalignedai.com


© 2022 Ethically Aligned AI Inc. All right reserved.