Sentencing Disparity Detector: A 2D Statistical Simulator |…

What the Sentencing Disparity Detector Is

The Sentencing Disparity Detector is an AI-ML based tool designed to simulate and analyze sentencing outcomes. It uses statistical methods, particularly z-scores, to flag cases that are statistically outside their peer group, suggesting potential bias in sentencing.

By linking three panels — a scatter plot of severity vs sentence, a bar chart showing per-judge disparity, and a histogram representing the population's z-score distribution — it provides a comprehensive view of how sentences vary within different groups.

Why It Happens

The phenomenon of sentencing disparity arises from various factors including individual biases, systemic issues, and differences in case severity. The detector helps to quantify these disparities by comparing each sentence against a statistical benchmark.

By tuning the z-score flag threshold, users can adjust how sensitive the system is to identifying outliers, which may indicate biased or unfair practices.

live demo · related simulation● LIVE

Real-World Applications

The principles behind the Sentencing Disparity Detector are used in real-world judicial-error and sentencing-disparity audits. These tools help legal systems identify areas where fairness might be compromised, leading to more equitable outcomes.

By understanding how these statistical methods work, policymakers and judges can make informed decisions about improving the justice system.

How It Works

The simulator generates a dataset of sentences based on severity, which is then analyzed using z-scores. Z-scores measure how many standard deviations an element is from the mean, allowing for a standardized comparison across different judges and cases.

By linking these panels, users can see in real-time how changes in sentencing patterns affect the overall distribution and identify potential biases.

Frequently asked questions

How does the z-score flag threshold work?

The z-score flag threshold determines at what point a sentence is considered statistically outside its peer group. Adjusting this threshold can help users focus on more or less severe cases of disparity.

What are some real-world examples of sentencing disparities?

Real-world examples include racial biases in sentencing, where minority defendants may receive harsher sentences than white defendants for the same crime. Another example is gender bias, where women might be sentenced more leniently or more harshly based on stereotypes.

Can this simulator detect all forms of bias?

While the simulator can highlight potential disparities, it cannot account for every form of bias that may exist in real-world sentencing. It is a tool to identify statistical outliers and suggest areas for further investigation.

How does tuning the number of cases affect the analysis?

Tuning the number of cases can change the overall distribution and z-score values, affecting how sentences are flagged as outliers. A larger sample size generally provides a more robust statistical analysis but may also include more noise.

Try it live

Everything above runs in your browser — open Sentencing Disparity Detector (2D) and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.

▶ Open Sentencing Disparity Detector (2D) simulation