Writing and Research

10/19/2020 OneNote https://uillinoisedu-my.sharepoint.com/personal/medhab2_illinois_edu/_layouts/15/Doc.aspx?sourcedoc={d1c5f23c-d3d7-4a70-9e0f-d9c219b4f0dd}&action=edit&w… 1/2 Summary of Sources Wednesday, April 15, 2020 10:06 PM It's not the algorithm, it's the data • Crime going down now, wanna keep this pattern • Predictive policing: likelihood that a person will commit another crime, end up in jail again • Moral, logistical, political issues: is it fair & ethical? • COMPAS(Correctional Offender Management Profiling)(owned by Northpointe. Inc) Actual and expected pretty close .71 Area Under Curve(AUC) • What about that 29%? Algorithm biased: relied on factors that could be correlated with race: poverty, postal codes, employment status Other places use it for a proactive approach Uses data from racially motivated or biased policing strategies • PredPol Uses 3 data points to provide predictions on where crime gonna occur: past type of crime, place of crime, time of crime Estimate risk • Motorola Solutions' CommandCentral Predictive Uses historical crime data and compares that with more recent data Allows changes to be seen in algorithms dIncorporates "concept of seasonality" to see hoe crime changes with temperature Expert(former police officer) says it's "a tool to help officers, not a replacement for the judgement of experienced officers" • Racially biased policing data gone into algorithms How can we change from these former mistakes? • Predicts crime in areas that have already known crime • Ignores non-reported crime "Estimated up to 15% of vehicle thefts and 65% of rapes or sexual assaults not reported" • But can provide a quick way to find trends and allocate resources there What is 'Fair'? Algorithms in Criminal Justice • Algorithms have been used in the criminal justice system since 1920s but now in pre-trial decisions • Judges look at risk scores determined by probability of rearrests and/or failure to appear in court Low risk scores: release defendants High risk scores: jail or bail(not affordable for all, can mean jail) • Northpointe's COMPAS ProPublica journalists called it racist Defendants who are not rearrested within 2 years but had high risk scores • 45% are black • 24^ white • Northpointe argues that within each risk category they have the same rearrests rate Historically black defendants are arrested more so they're more likely to get a higher risk score • Is it "ethical or constitutional to design and algorithm that treats people differently on the basis of race"(2) Richard Berk(statistician) calls it "matters of values and law…they are not matters of science"(2) Mathematically impossible to achieve "total fairness" in an algorithm Stakeholders have to determine the tradeoff • No algorithms inputs race, use inputs highly correlated with race and socioeconomic status • People in jail before their hearing are more likely to plead guilty • Questioning tradeoff between: harm that a person released commits another crime vs. holding a person in jail that wouldn't commit another one • "lack of transparency" Defendant in Wisconsin who sued state • "violated his rights to due process" because couldn't see the algorithm's code • Company says algorithm is propriety • Should defendants know their score and how they're calculated? • A lot of the times it's the judge deciding when the algorithms are used: not transparent on this But this decision can be random bc determined on an individual • Proposed requirements for pre-trial risk assessments Only used for certain conditions Tool transparent/ public should know inputs, outputs and the weight used Need to be check regularly ProPublica • Miscommunication cause Borden to be arrested for burglary and petty theft Had misdemeanors when she juvenile Black: rated high risk Never been charged with new crimes • Prater arrested for shoplifting abt 86 dollars of tools from Home Depo Had already been convicted of armed robbery and attempted armed robbery • Served 5 years in prison and another robber charge White: rated low risk Afterwards he was serving a 8 year prison from breaking into a warehouse and stealing thousands of dollars of electronics • Justice Department's National Institute of Corrections encourages combination of assessments • 2014 US attorney general Eric Holder said he fear the algorithms putting bias and wanted US Sentencing Commission to study it • Got risk scores from 7000+ people in Broward County, Florida Only 20% of people predicted to commit violent crimes actually did Blacks were more likely to be incorrectly marked high risk than whites Whites more likely to be incorrectly low risk than blacks • 137 questionnaire: pulled from records or filled out by person Asked opinionated questions too (disagree/agree) • Criminologists have been trying to make predictions abt crime since 1970s, now have the tech and ability to do so

RHET | 105 | Writing and Research | lecture notes | 10/19/2020 OneNote https://uil

10/19/2020 OneNote https://uillinoisedu-my.sharepoint.com/personal/medhab2_illinois_edu/_layouts/15/Doc.aspx?sourcedoc={d1c5f23c-d3d7-4a70-9e0f-d9c219b4f0dd}&action=edit&w… 1/1 Tech companies controlling minds Tuesday, January 21, 2020 9:31 PM • How to ethically steer someone's thoughts? • Notifications: schedules u to get sucked in it • Race for attention Meditation, games, social media • Youtube: autoplay videos Then netflix does it, then facebook • "Technology not neutral" • Snapchat: #1 way teenagers use to communicate Snapstreak: gives 2 people something they don't wanna lose, schedules blocks of time In 1970s telephones: didn’t have engineers who knew how your psyche worked • We don’t choose outrage Good for social media: gets attention better than calm feed "Accountable for maximizing attention" Precisely target a lie to most susceptible: profitable • "Taking away our agency to to spend our attention and live the lives we want" • 3 changes We are persuadable to protect our minds New models & accountability systems: so people are transparent to what people want; questioning business model of advisterising Design renaissance: protecting against timelines we don't want, empowering us to have the timeline we want • What's the best thing for you: host dinner party to talk abt controversial things • "superhuman ability" • Fix the way kids minds are to send empty minds • "Runaway artificial intelligence right now" • Can put our attention where we care about • "Addiction": interesting for them • "Impulsively" instead of what is the best use of your time • "uninted consequences" "lizard brain"

Courses in RHET

Popular Documents from Brigham Young University-Hawaii

Brigham Young University-Hawaii

Brigham Young University-Hawaii