SoylentNews is people
SoylentNews
Algorithms 'consistently' more accurate than people in predicting recidivism, study says:
In a study with potentially far-reaching implications for criminal justice in the United States, a team of California researchers has found that algorithms are significantly more accurate than humans in predicting which defendants will later be arrested for a new crime.
[...] "Risk assessment has long been a part of decision-making in the criminal justice system," said Jennifer Skeem, a psychologist who specializes in criminal justice at UC Berkeley. "Although recent debate has raised important questions about algorithm-based tools, our research shows that in contexts resembling real criminal justice settings, risk assessments are often more accurate than human judgment in predicting recidivism. That's consistent with a long line of research comparing humans to statistical tools."
"Validated risk-assessment instruments can help justice professionals make more informed decisions," said Sharad Goel, a computational social scientist at Stanford University. "For example, these tools can help judges identify and potentially release people who pose little risk to public safety. But, like any tools, risk assessment instruments must be coupled with sound policy and human oversight to support fair and effective criminal justice reform."
The paper—"The limits of human predictions of recidivism"—was slated for publication Feb. 14, 2020, in Science Advances. Skeem presented the research on Feb. 13 in a news briefing at the annual meeting of the American Association for the Advancement of Science (AAAS) in Seattle, Wash. Joining her were two co-authors: Ph.D. graduate Jongbin Jung and Ph.D. candidate Zhiyuan "Jerry" Lin, who both studied computational social science at Stanford.
More information:
Z. Lin, et al. The limits of human predictions of recidivism [open], Science Advances (DOI: 10.1126/sciadv.aaz0652)
(Score: 0) by Anonymous Coward on Wednesday February 19 2020, @11:56PM (4 children)
If you want to be technical
Heuristics - Something that generally, but not always, gets the right answer.
Algorithm - Something that always gets the right answer. It could be a mathematical formula used under the right circumstances such as the law of cosines
AI - More difficult to define but I think of it as something that seeks a desired outcome (probably using heuristics since it would get the desired outcome with algorithms each time) and if it doesn't get the desired outcome it tries to do more computations to figure out how it could get the desired outcome the next time it runs into the same situation and it stores some information to help it avoid getting an undesired outcome next time. It tries to find better answers for next time, or close approximations to the right answer.
An example could be a chess AI. The right answer would be a perfect move but it may require way too much computation. But if the computer sees that making a specific move under a specific condition causes it to lose it would try to figure out a better move and store the results for future games. It gets an answer that's closer to the perfect or 'right' answer, a better answer or an answer that more closely approximates a perfect answer (or the right answer).
(Score: 0) by Anonymous Coward on Thursday February 20 2020, @12:01AM
It tries to find better answers for next time, or closer* approximations to the right answer.
(Score: 0) by Anonymous Coward on Thursday February 20 2020, @12:08AM (1 child)
Let me delve into this a little bit more.
An example of a heuristic might be an antivirus that uses checksums to evaluate if something is a computer virus. These heuristics may not always be right but they can generally be right. But if the antivirus is wrong it will be wrong every time (if not updated), it doesn’t have a way to evaluate the outcome to search for a better solution next time.
A chess AI can evaluate the outcome. Did it win the game or lose. If it lost it can then try to do more computations and store some information to help it win the next time. For something to be intelligent it needs to be able to evaluate the outcome (ie: determine if it’s desired or not) and seek a different set of actions next time if the outcome is undesired so that it can get a desired outcome next time.
(Score: 0) by Anonymous Coward on Thursday February 20 2020, @04:12PM
Can't ya read the signs, boy? No loitering. No littering. No diving. No delving. Move along now.
(Score: 2) by FatPhil on Friday February 21 2020, @12:32AM
Nope. Totally utterly nope. That's so wrong I don't know where to start. It's probably not even wrong.
Here's my algorithm for working out the best move at chess given an input of a board position (plus ancillae):
1) If in check move out of check, with a preference to forwards over backwards, then left over right
2) If in check and the above failed, move the highest valued piece that can block in the way, tie-break on movement forwards, then leftwards
3) If in check and the above fail, capture the attacking piece with the highest valued piece that capture, tie-break on movement forwards, then leftwards
4) Else push the backmost outermost pawn that can move without discovering check by one forwards, with a preference of left over right
5) Else move the backmost outermost piece that can move without discovering or moving into check by the smallest possible (L_inf) distance, with a preference to not capturing over capturing, then forwards over backwards, then left over right
Precisely what do you think is "the right answer" about what it returns?
It's well defined, it's deterministic, and it always terminates with a suggested move, so it's most definitely an algorithm. (And most amazingly, I think it even follows the rules - I had to revisit it about 4 times to add more clauses.)
Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves