No, I don’t think that this study offers good evidence that installing air filters in classrooms has surprisingly large educational benefits. « Statistical Modeling, Causal Inference, and Social Science
> If there’s a problem, it’s with the general attitude in much of economics, in which it is assumed that identification strategy + statistical significance = discovery.
statistics  regression-discontinuity  false-discovery 
10 days ago
Berkeley Unified's students are coming from wealthier families — Berkeleyside
The population of Berkeley schools has become much more affluent over the past decade, with just over a quarter of students qualifying for free and reduced-price lunches — a proxy for tracking low-income students — compared to 44% in 2010-11, a new data set shows.

This year, only 26.7% of Berkeley Unified’s around 9,800 students can access the subsided school lunches, according to BUSD.
berkeley  gentrification  public-education  title-i  education  busd  berkeley-unified 
16 days ago
Examining Wedgewood: A Look at the Home-Flipping Giant in Battle With Homeless Mothers – NBC Bay Area
“By the end of the crisis, we saw from our count at least a million tenants in California were displaced by that activity of new owners coming in and giving tenants an eviction notice instead of allowing them to stay, and then flipping the properties.”
housing  speculation 
23 days ago
Defense Contractors Paid 'Protection Money' To Taliban, Lawsuit Alleges : NPR
In addition to making protection payments, one company — the telecommunications firm MTN Group — has been accused of deactivating its cellular towers at night at the request of the Taliban, which believed U.S. forces were using the cellular networks to track insurgents.
war  Afghanistan 
25 days ago
A Massive New Database Will Connect Billions of Historic Records to Tell the Full Story of American Slavery | History | Smithsonian Magazine
> . . . a massive new online database called “Enslaved: Peoples of the Historic Slave Trade,” which will launch in 2020. It aims to serve as a clearinghouse for information about enslaved people and their captors. Headquartered at Matrix, the Center for Digital Humanities & Social Sciences at Michigan State University, and funded by a founding $1.5 million grant from the Mellon Foundation, Enslaved will serve as a hub for many smaller digitization projects, Freedom Narratives among them.

> It turns out there’s a surprisingly simple way to turn life histories, ship manifests, census records and other information into machine-readable data: the semantic triple, which involves entering information in three-part sentences, each with a subject, a predicate and an object. “It’s something like, ‘Maria Picard was born in 1822,’ or ‘Maria Picard married Manuel Vidau,’” explains Rehberger.
slavery  history  archives  wikidata  triplestore  semantic-triple 
4 weeks ago
What Green Costs
> Getting the lithium entails sucking up the brine at an astounding rate. SQM, the company whose chairman I heard at the conference, pumps out brine at a rate of 1,700 liters a second — 95 percent of which is then evaporated. In other words, extracting lithium involves drawing up a lot of water and throwing most of it into the air.

> ... The brine water is underneath the salt flat. The freshwater systems are located at the flat’s perimeter. The two kinds of water are separated by a dynamic interface: a surface tension generated by the fluid’s differing density levels. Brine is much denser than freshwater, weighed down by dissolved elements like lithium. But while brine has the force of mass on its side, freshwater — which originates from snowmelt high up in the Andean peaks and the aquifers they feed — has the force of gravity in its favor. They are locked in a contest: mass versus gravity. When brine is removed, the interface separating them shifts towards the center of the salt flat, dragging freshwater with it — and away from the Indigenous communities located on the flat’s perimeter.

> Amid the overwhelming complexity of contemporary capitalism, it’s easy to forget that supply chains are not the product of geographic destiny. Indeed, a key aspect of environmental injustice is that contaminating processes — mines, power plants, or factories — are sited where ecosystems and human lives are seen as disposable or deemed to lack political influence.
lithium  battery  environment  environmental-justice  chile  argentina  bolivia  electric-cars 
5 weeks ago
Water Is Life: Nick Estes on Indigenous Technologies
> If you turn off the hydroelectric dam, the impact is catastrophic. The same goes for a nuclear power plant: if you’re not cooling your nuclear rods, there are disastrous downstream — literally, down the stream — consequences. You need hierarchical management built in, to keep people safe. But the existence of those threats is a manmade crisis that naturalizes and justifies that hierarchy once it’s been created.

> Our group, The Red Nation, dealt with this when we were planning a protest against the Entrada, a celebration of Spanish reconquest after the 1680 Pueblo Revolt that ultimately ended up being abolished by the city. We learned after the protest that the Sante Fe Police Department (SFPD) had issued a sealed warrant — so we never saw it at the time — and that Facebook had turned over all of our communications on Facebook Messenger. As a result, on the day of the action, the SFPD brought in twelve different law enforcement jurisdictions. There was a huge police presence with sniper nests and everything. Eight of our people got arrested. We didn’t find out until later that they had access to our Facebook messages.

> So demilitarization and carceral abolition are two main pillars of this program. We estimate that divesting from those state institutions would free up about a trillion dollars to reinvest in things like hospitals and healthcare and land that has been destroyed here, as well as in other countries that have been damaged by the US military.

> The Red Deal says that if we’re going to imagine carbon-free economies and the end of fossil fuels, then we also have to talk about decolonization. How are we going to build wind turbines but not give the land back to Indigenous people?

> There’s a Raytheon facility right outside of Farmington, New Mexico on Navajo land, where the workforce is 90 percent Navajo. That facility makes the microchips for the Israel’s Iron Dome missile defense system. The Alaska Native Corporation employs a lot of the private security forces who work at the child and family detention centers on the US-Mexico border. They also have contracts building what are essentially military bases in the Pacific. The Cherokee Nation has contracts to build State Department facilities in the Green Zone in Baghdad. There’s also a federal law that gives preference to Native businesses for lucrative defense contracts. These are the opportunities we get, and we have to take them because our subsistence economies have been annihilated . . . People say the Navajo Nation is dependent on coal and oil and gas, but I would actually say that the Southwest is dependent on the Navajo Nation producing coal and oil and gas because no one else wants to do it. No one else will have the generating station on their land because it’s one of the dirtiest coal-fired power plants in the country. So Navajo lands have been sacrificed — whether it’s been for coal, oil, and gas, or something like uranium. The same is true of Pueblo lands: the first atomic bomb was created on Pueblo land. And the nuclear waste that resulted was buried in Pueblo sacred sites because US government agencies knew Pueblo people would never tell anyone because they won’t say where their sacred sites are.
environment  environmental-justice  indigenous  sioux  standing-rock  climate-change 
5 weeks ago
US foster care problems lead kids to prison & homelessness | The Kansas City Star
> As part of its investigation, The Star surveyed nearly 6,000 inmates in 12 states — representing every region of the country — to determine how many had been in foster care and what effect it had on their lives.

> Of the inmates who took the survey, 1 in 4 said they were the product of foster care.
incarceration  foster-care  foster-care-to-prison-pipeline 
5 weeks ago
District Leaders Say Schools May Face $15-21 Million in Budget Cuts Next Year | Post News Group
> The Oakland Unified School District (OIUSD) has announced it may cut $15.5 million from its budget for the next school year, signifying there is no end in sight to proposed austerity measures including school closures that state officials have promised would lead eventually to a better day for Oakland students.

> . . . the Sacramento Unified School District is currently in the news as it struggles to come to grips with a $27 million deficit. The district is under pressure from FCMAT and the state auditor to . . . “[cut] salaries by 2 percent, [increase] teacher’s contributions to retiree health benefits and [cap] the district’s payment toward employee health care benefits at 90 percent.”
oakland  education  public-education  school-funding  labor  austerity 
5 weeks ago
The Racial Composition of Forensic DNA Databases by Erin Murphy, Jun Tong :: SSRN
Forensic DNA databases have received an inordinate amount of academic and judicial attention. From their inception, scholars, advocates, and judges have wrestled with the proper reach of DNA collection, retention, and search policies. Central to these debates are concerns about racial equity in forensic genetic practices. Yet when such questions arise, critics typically just assert that forensic DNA databases are not demographically representative. Such assertions are expressed in vague or conclusory terms, without a citation to actual data or even to concrete estimates about the actual composition of DNA databases.

This Article endeavors to fill these gaps in the literature by providing demographic information about the composition of forensic DNA databases. We draw upon two sources. First, we obtained data from states in response to our requests under freedom of information laws. Second, we devised an original estimate based on public information about each state’s DNA collection policies and the demographic data that matches those policies. In other words, we reverse-engineered the national DNA database.

We then use our data on the actual and estimated racial composition of DNA databases to identify and illuminate four questions fundamental to forensic DNA policy. First, the data center racial justice concerns as critical to debates about the proper scope of collection and search policies, as well as the impact of forensic DNA database practices more generally. Second, the data cast light on the significance, determinacy, and stability of race and ethnicity as meaningful biological and social categories. Third, the data provide insight into the advantages and disadvantages of choosing among architectural approaches when collecting, storing, and searching sensitive data such as genetic profiles. And finally, the data prompt questions about genetic privacy more generally, including how to weigh the significance of criminal justice practices in an increasingly genetically transparent society.
police-data  dna  dna-data  forensics 
9 weeks ago
Arrests as Regulation | Stanford Law Review
For some arrested individuals, the most important consequences of their arrest arise outside the criminal justice system. Arrests alone—regardless of whether they result in conviction—can lead to a range of consequences, including deportation, eviction, license suspension, custody disruption, or adverse employ­ment actions. But even as courts, scholars, and others have drawn needed attention to the civil consequences of criminal convictions, they have paid relatively little attention to the consequences of arrests in their own right. This Article aims to fill that gap by providing an account of how arrests are systemically used outside the criminal justice system. Noncriminal justice actors who rely on arrests—such as immigration enforcement officials, public housing authorities, employers, licensing authorities, and child protective service providers, among others—routinely receive and use arrest information for their own objectives. They do so not because arrests are the best regulatory tools but because they regard arrests as proxies for information they value, and because arrests are often easy and inexpensive to access. But when noncriminal justice actors rely on arrests, they set off a complicated and poorly understood web of interactions with the criminal justice system. Regulatory bodies and others that make decisions based on arrests can coordinate and pool resources with prosecutors and police officers, achieving a level of enforcement that neither could achieve alone, or they can make decisions that undermine important aspects of the criminal justice process. This Article maps different regulatory interactions based on arrests and illustrates the need for greater oversight over how arrests are used and disseminated outside the criminal justice system.
arrest  criminal-justice  police-data  arrest-as-regulation 
9 weeks ago
Bail Reform and Risk Assessment: The Cautionary Tale of Federal Sentencing - Harvard Law Review
Abstract: Part I of this Note examines contemporary bail practices, recent reforms, and risk assessments’ promises and shortcomings. Part II discusses federal sentencing reform, which originally sought a more empirical approach to criminal justice but failed. Part III applies the lesson of sentencing reform to bail reform today. Despite endorsing empirical tools, legislatures are prone to interfering with the evidence that informs those tools or with the tools themselves. Even after reforms, system actors retain misaligned incentives to incarcerate too many people. Technocratic instruments like risk assessments may obscure but cannot answer tough, fundamental questions of system design. But recent pretrial reforms have shown early signs of progress. If risk assessments are paired with adequate safeguards, sustained reductions in incarceration and progress toward equal treatment may be possible.


> Since the mid-1990s, “[f]ew outside the federal commission would disagree that the federal guidelines have been a disaster.” Two central problems have plagued the guidelines. First, federal sentencing failed as an expert, evidence-based endeavor. Almost from the get-go, Congress began to override the evidence-based reform goals of the guidelines by imposing congressionally determined guideline ranges and mandatory minimum sentences. Second, federal prosecutors have had too much power and leverage — both in individual cases and with sway over Congress — and have had no oversight and too few checks on that power.
bail  bail-reform  sentencing  criminal-justice  pretrial-detention  pretrial-justice  risk-assessment 
9 weeks ago
The Captured City — Real Life
> Contrary to the suggestions of “smartness” shills, these systems are not used by the general public but on it. This urban war machine (as I call it in my forthcoming book Too Smart) is the true essence of “smart” urbanism. It is the next step in the high-tech militarization of society. Rather than produce the smart city, it yields the captured city.
surveillance  internet-of-shit  internet-of-things  smart-cities 
10 weeks ago
[1905.05337] Scaling Bayesian Probabilistic Record Linkage with Post-Hoc Blocking: An Application to the California Great Registers
This paper uses Alameda County voter registration data from Alameda County between 1932 and 1936 as a record linkage case study. There are various theories about which demographic groups were most responsible for the large number of voters who changed parties during Roosevelt's first term, so this project seeks to test those theories against the Alameda County data by matching the records and then observing who changed parties (for this reason, they do not use party identification as one of the matching fields). The authors observe that, regardless of what model you're using (the paper uses the Fellegi-Sunter model with an additional penalty term in the likelihood function in order to control the number of true matches that are identified, and seeks to estimate the m- and u- probabilities), having a full posterior distribution over the parameters is preferable to, say, just having maximum likelihood estimates in order to characterize your uncertainty about the downstream inferences, and suggest sampling from a full Bayesian model for this reason. However, it's often too computationally intensive to do MCMC for anything but the smallest data sets. So this paper presents a sort of extension to blocking (not literally a type of blocking, as the authors point out). Note that the example in the paper is a case of "one-to-one" matching -- that is, the special case where you are matching two databases that are each assumed to be de-duplicated, so that each record from one data set has at most one matching record in the other one. From my reading, this constraint is not necessary for the logic they've laid out (the technique seems independent of the model you're using), but the effeciency gains they observe that makes it possible to run MCMC on larger data sets depends on this constraint, because this is the constraint that allows them to partition the overall record-linkage matrix into smaller blocks that can be optimized in parallel. So, anyways, the method: start by doing a coarse round of blocking using whatever method, then use that blocked data set to estimate match weights using a cheaper technique (e.g. expectation maximization for the Fellegi Sunter model). Then use these match weights to further filter the set of record pairs, by setting a threshold and zeroing out any weights that are less than the threshold. Having done that, you end up with a bunch of connected components that can themselves be treated as blocks, and since there can't be links between records in different blocks (because of the one-to-one constraint), you can make the decisions about which links are true within each block in parallel. That's the basic idea, as far as I can tell. There were a few sections of this paper that I had to re-read several times because I'm not sure I understood them properly, and perhaps that has to do with my interpretation of the method as "just" an additional filtering step beyond blocking in order to reduce the size of the set of candidate pairs. For instance, the authors claim they are able to use all blocked pairs in order to inform their inferences, rather than just those that make it through the weighing/thresholding step. I'm still working out how that works, and suspect I'll not fully understand until I've played with some code. To that end, this paper has an associated [Julia package](https://github.com/brendanstats/BayesianRecordLinkage.jl)
record-linkage  blocking  bayesian  probabilistic-record-linkage  julia 
11 weeks ago
Under digital surveillance: how American schools spy on millions of kids | World news | The Guardian
Some proponents of school monitoring say the technology is part of educating today’s students in how to be good “digital citizens”, and that monitoring in school helps train students for constant surveillance after they graduate.

“Take an adult in the workforce. You can’t type anything you want in your work email: it’s being looked at,” Bill McCullough, a Gaggle spokesperson, said. “We’re preparing kids to become successful adults.”
education  privacy  surveillance 
11 weeks ago
Philadelphia police are searching more cars for marijuana — but finding less of it, critics say
> In the last five years, police listed the smell of weed as the reason for more than 25,000 car stops and a growing number of searches. In the first quarter of 2019, searches related to the smell of marijuana surged to the highest level at least since 2014. Police identified the odor in 3,300 searches in that quarter, a number that has increased tenfold.

> At the same time, the “hit rate” — or rate at which contraband was found — plummeted to 9.7%, down from 40% in 2014.

> African American drivers represented 84% of those searched after an officer reported smelling marijuana. But those searches turned up marijuana only 12.6% of the time, compared with 20.3% of the time for white drivers.
discrimination  police-data  police-stops  philadelphia 
12 weeks ago
Incomplete and Garbled EOIR Court Data Suggest Lack of Commitment to Accuracy
> TRAC recently discovered gross irregularities in recent data releases from the Executive Office for Immigration Review (EOIR), the agency that oversees the US immigration court system.

> It is deeply troubling that rather than working cooperatively with TRAC to clear up the reasons for these unexplained disappearances, the agency has decided to dig in its heels and insist the public is not entitled to have answers to why records are missing from the data EOIR releases to the public.
data-quality  administrative-data  immigration  immigration-data  court-data 
12 weeks ago
The Body Instrumental
> Any measurement system, once it becomes integrated into infrastructures of power, gatekeeping, and control, fundamentally changes the thing being measured. The system becomes both an opportunity (for those who succeed under it) and a source of harm (for those who fail). And these outcomes become naturalized: we begin to treat how the tool sees reality as reality itself.

> When a technology assumes that men have short hair, we call it a bug. But when that technology becomes normalized, pretty soon we start to call long-haired men a bug. And after awhile, whether strategically or genuinely, those men begin to believe it.
gender  data-science  algorithmic-discrimination  schema-design  database 
12 weeks ago
NYPD Has Databases of Expunged Stop-And-Frisk Arrests, Bronx Defenders Say | The Marshall Project
“It’s not just Facebook and Google that have big data, it’s also police departments around the country using it to train the spotlight of their suspicion,” said Jenn Rolnick Borchetta, the lead attorney on the Bronx lawsuit. “But they’re running their algorithms and their facial-recognition software on arrest records and mugshots that were supposed to have been destroyed.”
police-data  police-misconduct  sealed-records  privacy  algorithmic-bias  zombie-predictions  administrative-data 
12 weeks ago
Dissecting racial bias in an algorithm used to manage the health of populations | Science
At a given risk score, Black patients are considerably sicker than White patients, as evidenced by signs of uncontrolled illnesses. Remedying this disparity would increase the percentage of Black patients receiving additional help from 17.7 to 46.5%. The bias arises because the algorithm predicts health care costs rather than illness, but unequal access to care means that we spend less money caring for Black patients than for White patients.
algorithmic-bias  health-care  health  bias  problem-formulation  labeled-data  label-bias  fairness 
october 2019
'One of the biggest, baddest things we did': Black Panthers' free breakfasts, 50 years on | US news | The Guardian
Last Saturday morning at Lil Bobby Hutton Park in West Oakland, local residents lined up for the 50th anniversary of the Black Panther party’s pioneering free breakfast program.
black-panther  black-history  oakland  west-oakland  free-breakfast 
october 2019
The Future of Warehouse Work: Technological Change in the U.S. Logistics Industry | Center for Labor Research and Education
> The warehousing industry is characterized by slim profit margins and cost-sensitive competition, which leads to a cautious approach to technology adoption.

> E-commerce is driving experimentation with new technologies.

> Technology uptake likely will be uneven.

> Technology potentially will have large impacts on third-party logistics firms and outsourcing in the warehousing industry.

> New technologies are likely to lead to work intensification.

> New technologies have the potential to de-skill some jobs.

> New technologies are poised to transform how workers are managed.

> In the short to medium term, new technologies likely will not cause widespread job loss.

> Technology is likely to have uneven impacts across demographics and occupations.
robots  workplace-issues  labor  technology  warehouse-work  automation  surveillance  worker-surveillance 
october 2019
Racial profiling analysis: LAPD searches blacks and Latinos more often - Los Angeles Times
> The Times analysis found that across the city, 24% of black drivers and passengers were searched, compared with 16% of Latinos and 5% of whites, during a recent 10-month period.

> Yet whites were found with drugs, weapons or other contraband in 20% of searches, compared with 17% for blacks and 16% for Latinos. The totals include both searches of the vehicles and pat-down searches of the occupants.

github link: https://github.com/datadesk/ripa-analysis
police-data  police-stops  police-misconduct  racial-profiling  stop-and-frisk 
october 2019
Closing Rikers: Competing Visions for the Future of New York City’s Jails | by Anakwa Dwamena | NYR Daily | The New York Review of Books
> As of two weeks ago, by the city’s count, there were approximately 7,000 people held across all of New York City’s jails, the majority of whom were at Rikers, at the cost of two billion a year, or around $300,000 per inmate. About 55 percent of the city’s incarcerated are black and 33 percent Latinx, even though only 24 percent of the city is black and 29 percent Latinx. Nearly half of the inmates have some form of mental health issue.

> Given that we know the highest risk factor for incarceration is contact with the criminal justice system, the best way to reduce the number of people imprisoned is not to incarcerate them in the first place. Of 7,000 people in New York City jails in April of this year, almost 5,000 were held pretrial; by some estimates, 40 percent of these would have been released under the New York State bail reform set to kick in at the beginning of next year.

> More than three quarters of people currently in New York jails have not been convicted of a crime and are awaiting trial; of these, 5 percent are in for a misdemeanor or lesser infraction, 22 percent for a nonviolent felony, 38 percent for a violent felony, and 11 percent for a suspected parole violation. Of the remainder, some 8 percent are locked up for an actual parole violation, 12 percent have been sentenced, and 5 percent are in transit or finishing out sentences. The only category of inmate population that is increasing is that of parole violators, a regulation enforced not by the city but by the State Department of Probation; this sanction is under consideration for reform in the New York State legislature as the current statute includes unjust stipulations like punishing violators not by the severity of the infraction but by their original crime.

> ... in 2017, 46 percent of all cases charged as violent felonies in the five boroughs were dismissed or resulted in an acquittal

> Eisenberg-Guyot argued that such abuses are baked in—“the ultimate endpoint of all jail and prison construction: conditions degrade because the dehumanization and violence is the point.”

> Here was the fundamental divergence laid bare: activists distrust the system, and see investment in people as a way to make it obsolete. The city distrusts the people and sees investment in the system as a way to keep them in check.

> Two weeks ago, the president of the Ford Foundation and a member of the Lippman Commission warned against taking an “extreme” position in the Rikers debate and letting “the perfect be the enemy of progress.” Four days later, speaking at Riverside Church, Angela Davis encouraged activists to show up outside the Ford Foundation building in protest, arguing that abolitionism is not an extreme position. Members of grassroots organizations like Decolonize This Place, Take Back the Bronx, and No New Jails showed up to the protest outside the foundation and about 300 current and former Ford Foundation fellows signed a letter in defense of abolitionism as a reasonable aim.
jail  pretrial-detention  abolition  prison-abolition  rikers-island  new-york 
october 2019
Arkansas' Phillips County remembers the racial massacre America forgot | Facing South
> ... the Elaine Massacre, perhaps the single deadliest instance of racist violence against African Americans to ever take place in the United States. The murder of hundreds of black farmers, sharecroppers, and businesspeople — men, women, and children — took place in the cotton fields of Elaine and the surrounding rural towns of Phillips County in the Arkansas Delta in late September and early October of 1919.

> Constantly shortchanged by the white planters, black sharecroppers and tenant farmers formed the Progressive Farmers and Household Union, which had chapters in the Phillips County towns of Ratio and Hoop Spur as well as Elaine. As the price of cotton skyrocketed from 1915 to 1919 and black landownership rates increased by 40 percent, the union determined it was time to demand fair prices and hired a white lawyer to take the case against the planters to court.
racism  violence  arkansas  jim-crow  mass-graves  unions  history 
october 2019
Nation’s Largest Bail Fund Plans To Stop Bailing People Out Of Jail - The Appeal
> Peter Goldberg, the group’s executive director, told The Appeal that he began to think about changing strategies when New York passed a bail reform package in June that did not end cash bail but instead established regulations for bail funds and effectively codified them as a permanent part of the carceral system.
pretrial-detention  pretrial-justice  bail  bail-reform  new-york  criminal-justice 
october 2019
Selection bias and the statistical patterns of mortality in conflict
This paper explores how information is generated about killings in conflict, and how the process of generation shapes the statistical patterns in the observed data. The difference between the observed patterns and the true patterns is called bias, two examples of which will be examined. First, we compare multiple individual sources reporting identifiable killings in Syria, highlighting variations in the likely probabilities of reporting for events of different sizes. Second, we conduct a similar analysis examining the number of sources reporting events of varying sizes in the Iraq Body Count public dataset. In both cases we explore how depending on the observed data without accounting for bias caused by missing data could mislead policy. The paper closes with recommendations about the use of data and analysis in the development of policy.
statistical-modeling  statistics  mse  multiple-systems-estimation  selection-bias  iraq  syria  event-size-bias 
september 2019
Data mining program designed to predict child abuse proves unreliable, DCFS says - Chicago Tribune
> A May 2017 Tribune investigation found the arrangement with Eckerd was among a series of no-bid deals Sheldon gave to a circle of associates from his previous work in Florida as a child welfare official, lawyer and lobbyist. Sheldon left Illinois under a cloud a month later, and a July joint report by the Office of Executive Inspector General and the DCFS inspector general concluded that Sheldon and DCFS committed mismanagement by classifying the Eckerd/Mindshare arrangement as a grant, instead of as a no-bid contract.

> The DCFS automated case-tracking system was riddled with data entry errors in both the Semaj Crosby and Itachi Boyle cases, the Tribune found. In addition, it did not link investigations about many children to cases regarding their siblings, or other adults in the same home.

> These and other shortfalls undermined Eckerd's analysis. And state laws forced DCFS to erase "unfounded" child mistreatment investigations, giving the Eckerd analysts less data to work with.

> The department is now moving to change the way it indexes and links investigations, and it is also considering legislative changes that would allow it to retain records of past unproven allegations.
automating-inequality  risk-assessment  surveillance 
september 2019
California bill looks to close data gaps in the criminal justice system | TechCrunch
> The California state legislature has passed AB 1331, a criminal justice data bill that aims to improve the quality of criminal justice records and creates a pathway for courts to share data with researchers.

> For example, the Department of Justice estimates that 60% of arrest records are missing disposition information, such as the judge’s ruling or sentencing. This can lead to criminalizing people who may be innocent. In addition, because pretrial risk assessment tools require timely and accurate information, any missing data could result in low-risk people being detained or high-risk individuals being released.
criminal-justice  arrest  police-data  pretrial-justice  risk-assessment  missing-data  measurement-bias 
september 2019
If You Give a Judge a Risk Score | The Little Dataset
I use this paper to show that, counter to expectations, the introduction of risk score recommendations can widen racial disparities for individuals who share the same predicted risk level.

This result is a consequence of two types of deviations by judges: across-judge and within-judge deviations. On the former, judges varied in their policy responsiveness; judges in whiter counties responded more to the new default (increasing their leniency) than judges in blacker counties. There is a striking correlation between a judge’s response to the policy and a judge’s defendant population. Second, even within judge and time, I show judges are more likely to deviate from the recommended default for moderate risk black defendants than for similar moderate risk white defendants. (Importantly, this is true in the post- but not the pre-period.) This result suggests that interaction with the same predictive score may lead to different predictions by race, which warrants further investigation.
pretrial-detention  pretrial-justice  risk-assessment  bias  fairness  discretion  judicial-bias 
august 2019
Risk of being killed by police use of force in the United States by age, race–ethnicity, and sex | PNAS
- [Risk of being killed by police use of force in the United States by age, race-ethnicity, and sex](https://www.pnas.org/content/116/34/16793). The main part of the paper goes over most of the same statistics presented in [the LA Times story](https://www.latimes.com/science/story/2019-08-15/police-shootings-are-a-leading-cause-of-death-for-black-men) about it, but in addition to those results, I was interested to read the supplemental material, which goes into the methods and some comparisons of the two datasets they used (the National Vital Statistics System and the Fatal Encounters database). For instance, by looking at trends over time in both datasets, they are able to hypothesize that under-reporting in Fatal Encounters decreased between 2000 and 2007 (perhaps due to more available online records). They discuss how missingness in the data can vary by group (race, sex, age), and how their approach to multiple imputation accounts for this, and how they simulated data in order to estimate the lifetime risk. They don't attempt to estimate the number of the unreported homicides. They use hierarchical models in order to calculate estimates for sparse strata (estimates by race + gender + age-group). Finally, at the end there is a discussion of inclusion criteria -- the authors included only cases where the classification made it clear that police use of force was the proximate cause of death. Notably, they exclude cases of suicide and various types of accidents. When they expand the inclusion criteria to include other types of deaths. As expected, the risk estimates increase, but they increase more for women than they do for men.
police-data  police-use-of-force  police-shooting  police-misconduct  hierarchical-models  multiple-systems-estimation  capture-recapture 
august 2019
Getting shot by police is a leading cause of death for U.S. black men - Los Angeles Times
- [Risk of being killed by police use of force in the United States by age, race-ethnicity, and sex](https://www.pnas.org/content/116/34/16793). The main part of the paper goes over most of the same statistics presented in [the LA Times story](https://www.latimes.com/science/story/2019-08-15/police-shootings-are-a-leading-cause-of-death-for-black-men) about it, but in addition to those results, I was interested to read the supplemental material, which goes into the methods and some comparisons of the two datasets they used (the National Vital Statistics System and the Fatal Encounters database). For instance, by looking at trends over time in both datasets, they are able to hypothesize that under-reporting in Fatal Encounters decreased between 2000 and 2007 (perhaps due to more available online records). They discuss how missingness in the data can vary by group (race, sex, age), and how their approach to multiple imputation accounts for this, and how they simulated data in order to estimate the lifetime risk. They don't attempt to estimate the number of the unreported homicides. They use hierarchical models in order to calculate estimates for sparse strata (estimates by race + gender + age-group). Finally, at the end there is a discussion of inclusion criteria -- the authors included only cases where the classification made it clear that police use of force was the proximate cause of death. Notably, they exclude cases of suicide and various types of accidents. When they expand the inclusion criteria to include other types of deaths. As expected, the risk estimates increase, but they increase more for women than they do for men.
police-data  police-use-of-force  police-shooting  police-misconduct  hierarchical-models  multiple-systems-estimation  capture-recapture 
august 2019
Technocolonialism: Digital Innovation and Data Practices in the Humanitarian Response to Refugee Crises - Mirca Madianou, 2019
Digital innovation and data practices are increasingly central to the humanitarian response to recent refugee and migration crises. In this article, I introduce the concept of technocolonialism to capture how the convergence of digital developments with humanitarian structures and market forces reinvigorates and reshapes colonial relationships of dependency. Technocolonialism shifts the attention to the constitutive role that data and digital innovation play in entrenching power asymmetries between refugees and aid agencies and ultimately inequalities in the global context. This occurs through a number of interconnected processes: by extracting value from refugee data and innovation practices for the benefit of various stakeholders; by materializing discrimination associated with colonial legacies; by contributing to the production of social orders that entrench the “coloniality of power”; and by justifying some of these practices under the context of “emergencies.” By reproducing the power asymmetries of humanitarianism, data and innovation practices become constitutive of humanitarian crises themselves.
technocolonialism  automating-inequality  palantir 
august 2019
Independent Analysis of State’s Attorney Data | The People's Lobby
Using data released by Kim Foxx’s office, clergy, legal experts and other advocates for criminal justice reform released a report on July 30, 2019 showing that sentences of incarceration have decreased by 19% during 2018, the Cook County State’s Attorney’s second year in office.
pretrial-justice  mass-incarceration  kim-foxx  district-attorney  police-data  court-data  incarceration 
july 2019
Receive report from the Office of the Public Defender relating to the establishment of a Pre-Arraignment Representation and Review (PARR) Team. - The County of Santa Clara, California
notes (and link to video) from the May 2019 Santa Clara Board of Supervisors meeting where funding for the Public Defender's Office PARR (pre-arraignment representation and review) program was approved.
sc-parr  court-data  court-watching  pretrial-detention  pretrial-justice 
july 2019
Silicon Valley De-Bug | The Future of Pretrial Justice is Not Money
> After court watching this process in felony court for months, seeing this same rerun, we decided to expand our “court watch” activities to “court doing” tactics — to not just observe, but to actively intervene and disrupt this machinery of detention. This is an expression of our participatory defense impulses for community to impact the court.

> We developed a community support identifying form with the public defender’s office who was also anxious to advocate for immediate release. The form is designed for those filling the courtroom pews, who want their support of their loved one facing charges to have value, a rightful place in the deliberation. The form asks them — someone’s aunt, friend, child, organization advocate — more about who their loved one is, the various ways they are part of this community, what’s at stake for them and others with another day of confinement, and also what roles they see themselves playing to assist that person in getting to court. Someone will say, “They can stay at my house,” or “I will give them a ride if they need it,” or explain how the person detained is who is responsible for taking the kids to school.
court-watching  pretrial-detention  pre-arraignment-defense  court-data 
july 2019
Metadata is the biggest little problem plaguing the music industry - The Verge
> That information [metadata] needs to be synchronized across all kinds of industry databases to make sure that when you play a song, the right people are identified and paid.

> Missing, bad, or inconsistent song metadata is a crisis that has left, by some estimations, billions on the table that never gets paid to the artists who earned that money.

> ... because there’s no standardized format for metadata, information often gets discarded or entered incorrectly as it’s written down or moved between people and databases . . . A label’s database is likely different from Spotify’s database, which is likely different from the databases of critical collection societies, like ASCAP and BMI, which pay public performance royalties to musicians.

> A song can pass through multiple songwriters, producers, and engineers before it gets released by an artist, and every new contributor adds the potential to screw things up. The longer the chain of custody for the data, the greater chance a portion of it will be incorrect.

> In 2016, the average hit song had over four songwriters and six publishers.
metadata  music  missing-metadata  administrative-data  standards 
june 2019
Aggression Detectors: The Unproven, Invasive Surveillance Technology Schools Are Using to Monitor Students
> By deploying surveillance technology in public spaces like hallways and cafeterias, device makers and school officials hope to anticipate and prevent everything from mass shootings to underage smoking. Sound Intelligence also markets add-on packages to recognize the sounds of gunshots, car alarms and broken glass, while Hauppauge, New York-based Soter Technologies develops sensors that determine if students are vaping in the school bathroom.

> Some experts also dispute the underlying premise that verbal aggression precedes school violence . . . “I can’t imagine when it would be useful, honestly,” said Jillian Peterson, an assistant professor of criminology and criminal justice at Hamline University in St. Paul, Minnesota.

> The software has been less effective at The Valley Hospital in Ridgewood, New Jersey. Daniel Coss, security chief for the hospital’s health system, said he’s phasing out the detector after a three-year, $22,000 pilot program. The devices — placed in public, “high risk” areas — had been set off by patients’ loud voices and cafeteria workers slamming cash registers closed. Once the detector was tweaked to be less sensitive, it ignored an agitated man who was screaming and pounding on a desk. The situation escalated until six security officers responded.

> During our first round of testing, when pizzas were delivered for lunch in the Sinatra library, the cheering triggered the detector. So did each round of Pictionary as students shouted guesses — ”A fireman!” “Lucifer!” — until the artist revealed the correct answer (Burning Man, the festival in remote Nevada). Laughter sometimes set it off, especially raucous guffaws that the detector apparently mistook for belligerent shouts.
surveillance  school-surveillance  school-to-prison  automated-behavior-detection  facial-recognition  speech-recognition  aggression-detection 
june 2019
New Jersey high school tests new technology to prevent school shootings | abc7ny.com
To decrease response times during school shootings, Rancocas Valley Regional High School in Mount Holly is using ZeroEyes, a system that automatically detects the presence of a gun from live video feeds. However, it needs to be trained on location first:

> For six months, the drills have been playing out every single week at Rancocas Valley Regional High School in Mount Holly: A gunman gaining entrance to the school and walking the halls.

> The gun isn't real, and the so-called "gunman" is the CEO of a new company called ZeroEyes.
predictive-policing  facial-recognition  surveillance  automated-firearm-detection  automated-behavior-detection 
june 2019
San Francisco says it will use AI to reduce bias when charging people with crimes - The Verge
starting July 1, the DA in San Francisco will begin using a tool built by the Stanford Computational Policy Lab that removes names of individuals (including the arresting officer) and identifiers of location, race, hair/eye color, etc. from the police arrest reports before the DA makes their charging decisions. The hope is that this will reduce bias in the charging stage. There is also evidence of racial bias in charges from the initial booking, which would not be addressed by this system.
police-data  district-attorney  racial-profiling  bias  named-entity-recognition  nlp 
june 2019
A Dangerous Brain | The Marshall Project
> Now a group of neuroscientists at the University of New Mexico propose to use brain imaging technology to improve risk assessments.

Researchers use "brain-age" (based on volume and density of gray matter) as an improvement over chronological age as a risk factor for recidivism.

> After analyzing the brain scans of 1,332 New Mexico and Wisconsin men and boys — ages 12 to 65 — in state prisons and juvenile facilities, the team found that by combining brain age and activity with psychological measures, such as impulse control and substance dependence, they could accurately predict rearrest in most cases.

experts disagree about how these predictions should be used (defense can use it as a mitigating factor, for example), the common debate about "risk assessment" vs. "needs assessment" -- apparently you can "encourage growth in brain areas linked to skills like empathy or self-control".
risk-assessment  neuroscience  phrenology  recidivism 
june 2019
Big Data Surveillance: The Case of Policing - Sarah Brayne, 2017
Tracks the intertwined growth of surveillance and "big data" within the Los Angeles Police Department through interviews and observation. Identifies five areas of transformation:

1. discretionary risk assessments -> quantified risk assessment: data that is input to these assessments includes the results of "field interviews," creating a feedback loop since "high risk" individuals are stopped for interviews more frequently, but each police contact adds to a person's risk score

2. reactive/explanatory analysis -> predictive analysis (e.g. PredPol)

3. query-based -> alert-based systems: the main example here is Palantir, which allows you to set alerts based on individuals, locations, or other characteristics present in real-time (often high-frequency) databases. Queries themselves become data, as the fact that someone has been searched by other officers using the system can itself be flagged

4. lower database inclusion thresholds: law enforcement databases have expanded beyond individuals who have direct contact with police through arrests or stops -- they now include data collected during stop-and-frisk and risk-based field interviews, the field interview data can include information on who else was with the person of interest even though they did not have any direct police contact, automated license plate readers (ALPR) suck up data constantly

5. integration of different data systems -- merging data across data stores and creating unique identifiers across systems, which governments might be interested in for the purpose of improving service delivery, also transforms the nature of surveillance -- interviewees rave about their Palantir software which lets them see everything in one place. In addition to data from public agencies including law enforcement, social services, health/mental health services, child/family services, the paper also mentions Palantir's constant inclusion of new data sources -- repossession/collections agencies, social media, foreclosure, electronic toll data, utility bills, pay parking lots, fast food call data, university camera feeds, rebate data . . . "In some instances, it is simply eaasier for law enforcement to purchase privately collected data than to rely on in-house data because there are fewer constitutional protections . . .". Much of the newly integrated data suffers from related types of inclusion bias (e.g. your chances of appearing in stop-and-frisk data differs based on race and class, this is also true for usage of social health/family services, etc., and even the placement and usage of ALPRs is based on measured crime rates), so that in all, these systems come to define and mark a population as suspicious (the only responses to queries will be people already in the data in some way).
surveillance  machine-learning  big-data  police  police-data  palantir  lapd  los-angeles 
june 2019
Looking Beyond the Visible Scene (PDF)
A common thread that ties together many prior works in scene understanding is their focus on the aspects directly present in a scene such as its categorical classification orthe set of objects. In this work, we propose to look beyond the visible elements of a scene; we demonstrate that a sceneis not just a collection of objects and their configuration or the labels assigned to its pixels - it is so much more. From a simple observation of a scene, we can tell a lot about the environment surrounding the scene such as the potential establishments near it, the potential crime rate in the area,or even the economic climate. Here, we explore several of these aspects from both the human perception and computer vision perspective. Specifically, we show that it is possible to predict the distance of surrounding establishments such as McDonald’s or hospitals even by using scenes located far from them. We go a step further to show that both humans and computers perform well at navigating the environment based only on visual cues from scenes. Lastly, we show that it is possible to predict the crime rates in an area simply by looking at a scene without any real-time criminal activity. Simply put, here, we illustrate that it is possible to look beyond the visible scene.
smart-cities  predictive-policing  google-street-view  image-recognition 
june 2019
A cryptoweaving experiment | Weaving codes – coding weaves
> . . . a woven artifact encodes time digitally, weft by weft. In most other forms of human endeavor, reverse engineering is still possible (e.g. in a car or a cake) but instructions are not encoded in the object’s fundamental structure – they need to be inferred by experiment or indirect means. Similarly, a text does not quite represent its writing process in a time encoded manner, but the end result. Interestingly, one possible self describing artifact could be a musical performance.
knitting  textiles  computation  history-of-mathematics  computer-history 
june 2019
Missing Numbers
Missing Numbers is a blog about the gaps in government data.
public-data  data  administrative-data  missing-data 
june 2019
We thought the Incas couldn't write. These knots change everything | New Scientist
A lost language encoded in intricate cords is finally revealing its secrets – and it could upend what we know about Incan history and culture
history  writing  khipu  history-of-mathematics  math  textiles  database  knots 
june 2019
Evelyn J Lamb – Inka History in Knots (Book Review)
Khipus record information in ways I found surprising. Aside from the numbers of knots and their positions on a khipu, the color, direction of twisting, and perhaps even fiber composition of the cords may also have had meaning. The concept of duality, which was important to the Inka people, may be encoded in khipu. If and when we are able to interpret narrative khipu, they may record stories or histories in ways that are completely unlike European linear, chronological narratives.
khipu  history-of-mathematics  math  south-america  inka  inca 
june 2019
Ferguson protesters dying: Six men with ties to Michael Brown protests have died since 2014 - CBS News
> Deandre Joshua's body was found inside a burned car blocks from the protest . . . The 20-year-old was shot in the head before the car was torched.

> Darren Seals, shown on video comforting Brown's mother that same night, met an almost identical fate two years later . . . The 29-year-old's bullet-riddled body was found inside a burning car in September 2016.

> MarShawn McCarrel of Columbus, Ohio, shot himself in February 2016 outside the front door of the Ohio Statehouse, police said. He had been active in Ferguson.

> Edward Crawford Jr., 27, fatally shot himself in May 2017 after telling acquaintances he had been distraught over personal issues, police said. A photo of Crawford firing a tear gas canister back at police during a Ferguson protest was part of the St. Louis Post-Dispatch's Pulitzer Prize-winning coverage.

> In October, 24-year-old Danye Jones was found hanging from a tree in the yard of his north St. Louis County home. The death was ruled a suicide. His mother, Melissa McKinnies, who was active in Ferguson, said it was a lynching.

> Bassem Masri, a 31-year-old Palestinian-American who frequently livestreamed video of Ferguson demonstrations, was found unresponsive on a bus in November and couldn't be revived. Toxicology results released in February showed he died of an overdose of fentanyl.
Ferguson  BlackLivesMatter 
june 2019
A shrug from voters to struggling L.A. schools, and have a nice summer, kids - Los Angeles Times
> “I think as LAUSD has become so heavily minority, so heavily poor … the public feels it doesn’t have a stake in public education anymore, and they’re willing to let conditions deteriorate,” said Sacks, whose class sizes are as high as 41 students.
lausd  education  public-schools  segregation  school-funding  taxes  california  los-angeles 
june 2019
To Track Police Conduct, a Flourishing of Data
> Public websites making police records accessible are central to the transparency movement, and over the past several years, nonprofit organizations, public defenders, academics, and journalists have mounted efforts like the Citizens Police Data Project, a large database of civilian complaints in Chicago, and CAPstat from the Cop Accountability Project, which contains data on federal civil rights lawsuits against police in New York City. Advocates in New York State, meanwhile, are pushing to repeal law 50-a, which allows police departments to shield misconduct records from disclosure. It is considered by many transparency advocates to be one of the most secretive such laws in the country.

> There are also similar projects underway, including a database from USA Today, to which the Invisible Institute contributed, which lists records of officers that have had their police certification removed; the Stanford Open Policing Project on traffic stops; and the Police Crime Database out of Bowling Green State University, which includes data on officers charged and convicted with crimes.
police-data  police-misconduct 
june 2019
The U.S. and China: A Tale of Two Surveillance States - The Atlantic
> Nelson Management is not the Chinese Communist Party. And Atlantic Towers is not Xinjiang. But in both places, minorities are being surveilled with nascent technology and subjected to an organized effort to control their behavior.

Comparing surveillance in Xinjiang and in the Atlantic Plaza Towers in East Brooklyn.
surveillance  facial-recognition  xinjiang  china  automating-inequality 
june 2019
Immigrants In El Paso Were Held In Dangerously Overcrowded, Standing Room–Only Cells
> At a processing center in El Paso, the Department of Homeland Security’s Office of Inspector General (OIG) found that facility didn’t have the capacity to hold the hundreds of people inside. El Paso del Norte Processing Center’s maximum capacity is 125 detainees, but on May 7 and 8, Border Patrol’s custody logs showed there were about 750 and 900 detainees, respectively, which is counter to the agency’s standards.
migration  immigration  border  border-patrol  cbp  texas  el-paso  detention  cruelty 
june 2019
At this L.A. supper club, refugees share food and memories of the lives they left behind - Los Angeles Times
> ... New Arrivals Supper Club, a communal feast prepared by immigrants from some of the world’s most troubled countries. Launched in 2017 by the L.A.-based nonprofit Miry’s List, the monthly dinners — hosted in homes and restaurants — aim to empower newly settled refugees by giving them opportunities to earn money, forge new communities and share their culture through food.
refugee  immigration  los-angeles  kindness  supper-club 
may 2019
Lockport Public Schools Will Be The First In The Country To Use Facial Recognition On June 3
(after publication of the story, the NY Dept of Education asked Lockport to delay the rollout of the facial recognition, as the Department is in the process of drafting standards for data privacy and security, and would like the district to align the program with those regulations once they are out).

> In March 2018, Lockport announced its plans to install a facial recognition security system, which it funded through the New York Smart Schools Bond Act — an act meant to help state schools augment their instructional tech. But instead of buying laptops and iPads, Lockport submitted a proposal for a high-tech security system, and allocated much of the $4.2 million it was given toward adding dozens of surveillance cameras in the school and installing the facial recognition system Aegis, which is provided by Canada-based SN Technologies. To date, Lockport has spent $1.4 million to get the system up and running.

> Aegis will track individuals who are “level 2 or 3 sex offenders, students who have been suspended from school, staff who have been suspended and/or are on administrative leave, any persons that have been notified that they may not be present on District property, anyone prohibited from entry to District property by court order … or anyone believed to pose a threat based on credible information presented to the District.” The Lockport Journal further reported that the object recognition system will also be able to detect 10 types of guns.
facial-recognition  educational-technology  education  public-schools  privacy  student-privacy  school-surveillance  school-to-prison 
may 2019
Workers Should Be in Charge
The Democracy Collaborative report proposes that companies being sold or closed would be held in escrow for a period of time. Workers would be given the right to choose a trustee, or have one appointed on their behalf, and would be made aware of how much they needed to pay to exercise the right of first refusal. They would then be given access to a range of new dedicated sources of capital for worker ownership transitions, including ones that would mandate repayment not in the form of money but instead by addressing social needs through production, changing company practices to be more environmentally sustainable, and/or rectifying legacies of discrimination or inequality in the workplace.
worker-owned  cooperative  labor 
may 2019
Kiwibots win fans at UC Berkeley as they deliver fast food at slow speeds - SFChronicle.com
> The Kiwibots do not figure out their own routes. Instead, people in Colombia, the home country of Chavez and his two co-founders, plot “waypoints” for the bots to follow, sending them instructions every five to 10 seconds on where to go.

> As with other offshoring arrangements, the labor savings are huge. The Colombia workers, who can each handle up to three robots, make less than $2 an hour, which is above the local minimum wage.

> On the ground in Berkeley, people also do a lot of robot support. Traveling at 1 to 1½ mph, the bots would take too long to chug to local restaurants, so Kiwi workers pick up the food at restaurants and take it via bikes or scooters to meeting spots around campus to insert into an insulated bag in the bots’ storage compartment.
artificial-intelligence  hype  robotics  outsourcing  labor  tech-workers 
may 2019
San Francisco police chief concedes raid on journalist was wrong — ‘I’m sorry’ - SFChronicle.com
Police asked Carmody to reveal his source in April. When he refused, officers showed up with a sledgehammer, battering ram and pry bar before seizing his computers, cameras and phones at his home and office. They handcuffed Carmody as well . . . Scott said the officers who executed the search did not consult with the district attorney’s office before obtaining the warrants
san-francisco  police  police-misconduct  1st-amendment  freedom-of-press 
may 2019
Depression-Gene Studies Were Built on Shaky Foundations - The Atlantic
When geneticists finally gained the power to cost-efficiently analyze entire genomes, they realized that most disorders and diseases are influenced by thousands of genes, each of which has a tiny effect. To reliably detect these minuscule effects, you need to compare hundreds of thousands of volunteers. By contrast, the candidate-gene studies of the 2000s looked at an average of 345 people! They couldn’t possibly have found effects as large as they did, using samples as small as they had. Those results must have been flukes—mirages produced by a lack of statistical power.
science  narratives-from-noise  hype  replication 
may 2019
Will controversy over risk assessments break Philly’s touted criminal-justice reform collaboration?
Three years into a MacArthur Safety & Justice Challenge grant, a promised "Risk Assessment Tool (RAT)" for determining pretrial release conditions is . . . well, I guess the question is where it is. The D.A.'s office and the Defender's Association (hereafter "DA/PD"), both interested in ending cash bail, feel there's been no progress on the tool or on ending cash bail, and have withdrawn their support for the RAT. Meanwhile Dr. Richard Berk, who's supposed to be developing the tool, says he's made a lot of progress, but no one (at least not among the withdrawal letter signatories) has seen any information beyond a list of factors that would be considered by the tool. A couple of tidbits from the story:

- "City and court leaders have emphasized that no major jurisdiction has eliminated the use of money bail without a risk tool." -- Is this true? I can't think of any counter-examples, but haven't seen the claim made before. Also, the Defender Association has proposed an alternative -- presumptive release for most charges with an arraignment hearing for exceptions (for reference, the D.A.'s office under Krasner has stopped seeking bail for a number of charges that together account for ~63% of charges -- though they do not control whether bail is set, they do influence it by their recommendations, as demonstrated in the Megan Stevenson paper that I wrote about a couple of weeks ago). The letter from the DA/PD claims their proposal was ignored.

- One complaint in the DA/PD letter is that ". . . risk had been defined as the likelihood of re-arrest in 18 months, a definition on which we were not consulted and with which we cannot agree. When pressed, [Dr. Jaime Henderson, director of reseearch and development for the courts] told those present that this definition had been created in 2012, at least three years before we applied for our MacArthur grant."
philadelphia  cash-bail  bail-reform  bail  pretrial-detention  risk-assessment 
may 2019
Flux – What Is Differentiable Programming?
> Differentiable programming applies the techniques of deep learning to complex existing programs, taking advantage of the huge amount of knowledge embedded within them.

> For example, ConvNets have a huge advantage over the perceptron because they re-use image kernels, which are long known to exploit translational invariance.

> As in the sciences, hybrid models can both be more effective and resolve some of the tradeoffs between deep learning and explicit programming.

> The final step taken by differentiable programming is to no longer see matrix multiplies, convolutions and RNNs as the fundamental building blocks of deep learning, but instead as mere special cases. We can apply the techniques of deep learning to any parameterised, differentiable function f(x)f(x)f(x). fff could be a matrix multiply, but functions as complex as physics simulators or ray tracers can be differentiated and optimised just as well. Even quantum computation can fit into this framework.
differentiable-programming  machine-learning  deep-learning  modeling 
may 2019
An algorithm wipes clean the criminal pasts of thousands - BBC News
> ... California’s Proposition 64, a measure passed in 2016 that made marijuana legal in the state. As part of the new law, those with prior convictions could now seek to have them struck off their record.

> “I formed the opinion that this is really our responsibility,” said George Gascon, San Francisco’s district attorney.

> [Code for America] made Clear My Record, a tool that can analyse text in court files, using character recognition to decipher scanned documents.
code-for-america  machine-learning  marijuana  expungement 
april 2019
Here Come AI-Enabled Cameras Meant to Sense Crime Before it Occurs - Defense One
> To teach the algorithm [that detects when a person enters a facility with a gun], a team from the company [ZeroEyes] shows up on location and proceeds to stage mock attacks.

> Athena [a Peter Thiel-backed startup] sells software that can recognize specific behaviors, like fighting, walking slowly when others are walking fast, virtually anything that aberrates from the norm.
predictive-policing  palantir  facial-recognition  surveillance  automated-firearm-detection  automated-behavior-detection 
april 2019
« earlier      
2nd-amendment activism administrative-data advertising affirmative-action affordable-housing ai algorithm algorithmic-bias amazon antifa antitrust archives art artificial-intelligence automating-inequality automation bail bash bay-area bayes bayesian bayesian-inference berkeley bias blacklivesmatter blocking border california capitalism car-culture cartography cash-bail causality census charter-schools child-detention civil-war climate-change code command-line confederacy copyright corporate-university corruption court-data crime-data criminal-justice criticism cruelty css d3js data data-analysis data-collection database datascience dataset dataviz ddj deathpenalty democracy density design detention development district-attorney drone-war drones economics economy edtech education elections environment ethics explanation facebook facial-recognition fairness fake-news family-separation fascism fatml ferguson fire first-amendment free-speech functional gender gentrification ggplot2 gig-economy gis google gui guns hci health health-care hierarchical-models history homelessness housing housing-policy housingsegregationineverything html human-rights hype ice immigration incarceration inequality infrastructure intellectual-property internet internet-of-things irs-990-data jail javascript jia-tolentino justice kkk labor labor-rights labor-unions language law-enforcement linguistics linux longreads los-angeles machine-learning machinelearning mapping maps marijuana math mathematics migration misogyny modeling music nazi neoliberalism neuralnets nimby nlp oakland opendata orms palantir pasta philadelphia philanthropy police police-data police-misconduct police-shooting police-use-of-force political-debate politics poverty prediction predictive-modeling predictive-policing pretrial-detention pretrial-justice prison privacy programming public-data public-housing public-infrastructure public-policy public-schools public-space public-transportation python r race racism recipe record-linkage refugee replication restaurants risk-assessment rmarkdown rstats san-francisco sb827 schema-design schools science science-communication search segregation sexism sexual-harassment sharing-economy silicon-valley slavery smart-cities social-justice sql stan statistical-modeling statistics strike surveillance surveillance-capitalism tax-policy tech tech-workers technology text-interface tidyverse tomato tools transportation trump twitter typography uber ucberkeley ui unions university unix urban-planning ux violence visualization war white-supremacists whiteness womeninstem writing yimby zoning

Copy this bookmark: