Our justice system requires that someone is found guilty of a crime only if the evidence presented at trial demonstrates beyond reasonable doubt that they did indeed commit the crime. Otherwise, they are found not guilty and hence innocent.
Unfortunately this simple principle is routinely ignored and undermined in our justice system when, for example:
We are a small, independent, family team who share a deep frustration with injustices in criminal cases that result in innocent people going to jail.
Simon Mitchell is a computational biologist who mainly works on cancer research at the University of Sussex. More about Simon and his group’s main work here: https://www.bsms.ac.uk/about/contact-us/staff/dr-simon-mitchell.aspx, and here: https://mitchell.science/
Cliff Mitchell and Jane Mitchell, who are both previously university-based scientists/researchers, with careers spanning medicine, industry, writing, mathematics, information technology and business, are now working on this project as independent researchers.
We as a family share a deep frustration with injustices in criminal cases that result in innocent people going to jail. When discussing this we realised our unique collection of skills could possibly do something about this issue, potentially through some of our computational experience and machine learning. So we worked together on writing a paper to figure out whether, if we used machine learning, we could accurately assign criminal cases a guilty or not-guilty verdict, even when the jury got it wrong. Remarkably it worked, and we published the paper in Law Probabiity and Risk (https://academic.oup.com/lpr/advance-article-abstract/doi/10.1093/lpr/mgaa003/5807887). We think this really could be a way to avoid people going to jail who shouldn’t.
The project relies on people following a trial in the same way a jury would, and inputting scores into the machine learning algorithm. For the paper we did this ourselves, however because we are a small family research team, we have a challenge in proving this approach could work in practice because:
Therefore it would be really transformative for this project if we could recruit even a handful of independent volunteers to watch some freely-available trial proceedings and input their scores into the algorithm. That work would result in a further publication that would provide the evidence necessary to really push for adoption of this approach in trials, in a way that could prevent miscarriages of justice and all the associated suffering.
justiceBRD is an attempt to bring our justice system back in line with its fundamental principle by forcing a decision to be based solely on the strength of the evidence presented at trial. We do this by focusing on murder trials, as it is here that miscarriages of justice are most profound and most damaging to individuals and society.
For each case, justiceBRD identifies the actual strands of evidence presented by the prosecution or defence, and makes an objective assessment of each strand through a series of challenging, standardised questions. The questions have been developed in order to avoid a range of known failings in the justice system, including those identified above. The answers to these questions are then entered into a maching learning model. This model has been 'trained' on a number of murder cases where the true outcome is already know and then tested on many other known cases and shown to produce the correct verdict with 100% accuracy. The output from the model is a verdict together with a degree of certainty that the accused is guilty, based on the actual evidence. That is whether or not the defendant is guilty beyond reasonable doubt.