liveBook · Manning

Risk assessment

As explained in Inside AI

Risk assessment is the process of evaluating the likelihood and potential impact of uncertain adverse events.

Risk assessment evaluates the likelihood and potential impact of uncertain adverse events. In criminal justice, numerical scores can inform bail, incarceration, sentencing, alternative sentences, and parole decisions. section 10.2

Algorithmic recidivism prediction

A ProPublica analysis examined more than 7,000 Broward County arrests from 2013–2014. The scores predicted violent offenses unreliably: only 20 percent of those flagged actually committed violent crimes. section 10.2

The example shows why risk assessments require scrutiny:

  • Predictive performance for broader offenses was only slightly better than chance. section 10.2
  • The formula flagged Black defendants as future criminals at almost twice the rate of white defendants. section 10.2

How Inside AI covers it

How deep it goes
Defines the practice through courtroom applications and outcome analysis, but stops short of explaining algorithm design.
Examples
Uses bail, sentencing, parole, and the 2016 ProPublica analysis of Broward County risk scores.
What you can do after
Evaluate predictive scores by comparing forecasts with later behavior and examining disparities across racial groups.
How it connects
Connects risk assessment with predictive algorithms, criminal justice, recidivism, and machine bias.

Also explained in

Compare how 4 books explain risk assessment →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Inside AI. The section links point to the parts of the book it draws on.