to me in sentencing is to look very uh is to have the conversation with the defendant not with the prosecutor not with the defendant 's lawyer but to be able to talk to the person I'msentencing and to look into his or her eyes and mostly it's his eyes but there are some female
sentencing and to look into his or her eyes and mostly it's his eyes but there are some female defendants uh and make sure that I am able to communicate to them what thissentence is and why and there has been at least there's been more than one but one case where I truly did not believe
Some homes are for a nonprofits. The defendant has an attorney.
Does that make sense? Black defendants , yes, have typically had higher recidivism rates because of historical deeply entrenched economic and sociopolitical inequality.
many different causes. So your brain has this dilemma. And it doesn't just feeling that the defendant can't be trusted. So, your brain is basically
So normally what we do is we take a machine learning algorithm. We predict the defendant 's risk. We do a horse race between the algorithm and the judge.
We do a horse race between the algorithm and the judge. So the defendant 's behavior is the object of interest. The other thing that we can do is basically turn the algorithm on the judge themselves.
And then what we've done is we've binned defendants up. We've grouped defendants together based on the algorithm's predicted defendant crime risk. So we are looking at the dispersion of judge's release probabilities as a function of how serious the defendant 's
And so you can see how my friend-- like the fact of having the snack was an enormously salient detail that turned out completely blinded to defendant race or job applicant race-- like, for sure that's true.
And then they would divide the class into halves randomly and one half would argue for the defendant and the other half would argue against the defendant . And then the two sides went off and studied the facts and then they made their arguments.
those things just worry me about our country and about access to education and jobs and transportation and health interviews the defendant family members teachers all sorts of people get
You have a trial. There's a defendant . Is that defendant going to be convicted and taken out the back of the courtroom in handcuffs and go to jail for decades? Or are they going to go out the front door a free person?
What had happened, the judge had taken a bribe from my co-defendant 's attorney to throw the case. My co-defendant 's attorney had been corrupt and had bribed the judge on other cases. I knew nothin' about the bribe.
During her trial, two Jews described as 'eminent jobbers in secondhand clothes', appear in court to testify. As one of them explains to the jury, they both married the defendant on the same day, because quote, "He thinks if we can get one wife between us, that will be less expensive.” The crowd no doubt roared in appreciation of the joke.
In particular, it was twice as likely to falsely predict future criminality in black defendants than white defendants . And the article provided many examples, one of which is right here on the slides. And you have an individual, a male who had committed two armed robberies, and he had a very low risk score in comparison
Been to witness? What sort of people do you see? Who are the defendants ? What type of people, if you were just broadly characterizing them?
So for example, the COMPAS criminal justice profiling system was sentencing black defendants far more frequently than white defendants into much more severe sentences, because it was using a data set that showed that black defendants were more likely to commit crimes again in the future.
If you've got a 10-year bid on your first day in the system, you should start to build what you need to do for reentry services upon your release, So any low-income defendants who may have trouble or somehow impeding transportation or any other inconvenience to get to court, my goal is to have a virtual hearing court
e.g., better recognition of dark-skinned faces." And what that is acknowledging right there is that that product went to market not as 48% of white defendants labeled low risk did re-offend, and 28% of black defendants re-offended.
become an incarcerator i conduct civil and criminal trials I take guilty pleas and I sentence defendants and do all of the miscellaneous duties that are required uh for a judge to do it's an awesome responsibility and one that I welcome but I don't always feel that I'm
Is it all right? against criminal defendants . And, since the police are the people who are out there selectively enforcing laws and disproportionately and aggressively
And so here's what the issue is, right? If the judge releases a defendant -- if the judge releases a defendant , we don't have any sort of problem. I can observe what the crime outcome is under the judge's actual decision, and I know what the crime outcome is under the counter-factual algorithmic decision
I can observe what the crime outcome is under the judge's actual decision, and I know what the crime outcome is under the counter-factual algorithmic decision if the algorithm wants to detain the defendant , because putting someone in jail, by definition, incapacitates them and prevents them from engaging in crime. So that's easy. But what if the judge detains the defendant ?
So that's easy. But what if the judge detains the defendant ? What if the judge detains the defendant , and the algorithm wants to release them? How do I know whether the algorithm is right or whether the judge was right, OK?
Here's the problem in a policy application, which is the judge sees things that no algorithm will ever see. So this is an example of a defendant , a 25-year-old guy who reported his occupation, when he was arrested, as tattoo model.
So here's what we're going to do. I got a fact pattern about a defendant . And I'm going to give this group this fact pattern, and I'm going to give this group this fact pattern.
it's like to be the defendant 's mother-- in this case, two mothers going through a lot of turmoil and heartache.
How do we assure a defendant a fair trial?
Upon finding me guilty, I really thought that the judge was just prejudiced against me. And I had a co-defendant as well. And I really thought the judge was just prejudiced against me.
So after all my co-defendants had accepted plea deals, miraculously I was offered a plea deal that if I accepted a first-degree felony,
I try to convince that defendants are prone to make inevitable choices due to their state of mind
security prison and so our it does not it has a medical facility but it's not for people with specific medical or mental health issues a lot of the defendants I sentence are dispersed all over the country and so it's not as easy as just driving to mil and and talking to people uh but I'm very interested in
prisoner re-entry and trying to get connected there's two judges in Detroit who are working on meeting prisoners uh defendants as they come back into the community and so it's my hope to spend some time with that project thank you
80% to 85% of criminal defendants are too poor to afford their own lawyers, much less an investigator or an expert
and he's gone from torturing defendants --verbally anyway-- to torturing readers.
Over 90 percent of the courts defendants were Eastern border smugglers and undocumented workers mostly from the neighboring town of Sonneberg.
He looked more like a hostage than a defendant .
bail, and parole decisions. So these algorithms would look at the profile of a defendant and then, based on historical data, come up with various risk scores-- for example, a recidivism score that would predict the likelihood that this person would reoffend.
Who do we find? Has anyone spent any time-- I don't mean as a defendant -- as an observer or something, or as a lawyer, even, in a Magistrates' Court or the lower courts of any country? Been to witness? What sort of people do you see?
On either side were the prosecutor and the defendant and the defense attorney.
This does not look like a very high-risk guy. The judge can see something extra about the defendant . Now, imagine what happens.
The first insight is to recognize that the problem's one-sided. So when the algorithm wants to release a jailed defendant , that's a problem, but there's no problem if the algorithmic decision rule wants to jail someone that the judge released.
And I'm going to give this group this fact pattern, and I'm going to give this group this fact pattern. I'm going to describe to you a defendant and all the evidence about that defendant . You get the same fact pattern.
those things just worry me about our country and about access to education and jobs and transportation and health incredible probation department that goes out to the defendant 's homes
Who claimed to be present. And actually held the woman down while me and my co-defendant stabbed her to death. Yet she was free to go home after giving that statement.
And so here on the left, we've got the judge with the 90% release rate. We can go into the pool-- the remaining 90% of defendants that the lenient judge has released-- and we can rank order defendants by our algorithm's predictions of their crime risk.
predictions of their crime risk. And then we can count down that rank-ordered list of defendants by predicted risk and pick another 10% of defendants to detain. So this is what the algorithm's guess is for the most socially productive marginal 10% of defendants to detain.
sorry, the judge's predicted detain probability on the y-axis. And then what we've done is we've binned defendants up. We've grouped defendants together based on the algorithm's predicted defendant crime risk.
e.g., better recognition of dark-skinned faces." And what that is acknowledging right there is that that product went to market not as So the software was particularly likely to falsely flag black defendants as likely to re-offend.
that the hairs didn't come from any of these defendants .