Report on Algorithmic Risk Assessment Tools in the U.S. Criminal Justice System

Curated on Posted on by Stefaan Verhulst

Press release: “The Partnership on AI (PAI) has today published a report gathering the views of the multidisciplinary artificial intelligence and machine learning research and ethics community which documents the serious shortcomings of algorithmic risk assessment tools in the U.S. criminal justice system. These kinds of AI tools for deciding on whether to detain or release defendants are in widespread use around the United States, and some legislatures have begun to mandate their use. Lessons drawn from the U.S. context have widespread applicability in other jurisdictions, too, as the international policymaking community considers the deployment of similar tools.

While criminal justice risk assessment tools are often simpler than the deep neural networks used in many modern artificial intelligence systems, they are basic forms of AI. As such, they present a paradigmatic example of the high-stakes social and ethical consequences of automated AI decision-making….

Across the report, challenges to using these tools fell broadly into three primary categories:

  1. Concerns about the accuracy, bias, and validity in the tools themselves
    • Although the use of these tools is in part motivated by the desire to mitigate existing human fallibility in the criminal justice system, this report suggests that it is a serious misunderstanding to view tools as objective or neutral simply because they are based on data.
  2. Issues with the interface between the tools and the humans who interact with them
    • In addition to technical concerns, these tools must be held to high standards of interpretability and explainability to ensure that users (including judges, lawyers, and clerks, among others) can understand how the tools’ predictions are reached and make reasonable decisions based on these predictions.
  3. Questions of governance, transparency, and accountability
    • To the extent that such systems are adapted to make life-changing decisions, tools and decision-makers who specify, mandate, and deploy them must meet high standards of transparency and accountability.

This report highlights some of the key challenges with the use of risk assessment tools for criminal justice applications. It also raises some deep philosophical and procedural issues which may not be easy to resolve. Surfacing and addressing those concerns will require ongoing research and collaboration between policymakers, the AI research community, civil society groups, and affected communities, as well as new types of data collection and transparency. It is PAI’s mission to spur and facilitate these conversations and to produce research to bridge such gaps….(More)”

LAPD moving away data-driven crime programs over potential racial bias

Curated on Posted on by Stefaan Verhulst

Mark Puente in The Los Angeles Times: “The Los Angeles Police Department pioneered the controversial use of data to pinpoint crime hot spots and track violent offenders.

Complex algorithms and vast databases were supposed to revolutionize crime fighting, making policing more efficient as number-crunching computers helped to position scarce resources.

But critics long complained about inherent bias in the data — gathered by officers — that underpinned the tools.

They claimed a partial victory when LAPD Chief Michel Moore announced he would end one highly touted program intended to identify and monitor violent criminals. On Tuesday, the department’s civilian oversight panel raised questions about whether another program, aimed at reducing property crime, also disproportionately targets black and Latino communities.

Members of the Police Commission demanded more information about how the agency plans to overhaul a data program that helps predict where and when crimes will likely occur. One questioned why the program couldn’t be suspended.

“There is very limited information” on the program’s impact, Commissioner Shane Murphy Goldsmith said.

The action came as so-called predictive policing— using search tools, point scores and other methods — is under increasing scrutiny by privacy and civil liberties groups that say the tactics result in heavier policing of black and Latino communities. The argument was underscored at Tuesday’s commission meeting when several UCLA academics cast doubt on the research behind crime modeling and predictive policing….(More)”.

Introducing the Contractual Wheel of Data Collaboration

Curated on Posted on by Stefaan Verhulst

Blog by Andrew Young and Stefaan Verhulst: “Earlier this year we launched the Contracts for Data Collaboration (C4DC) initiative — an open collaborative with charter members from The GovLab, UN SDSN Thematic Research Network on Data and Statistics (TReNDS), University of Washington and the World Economic Forum. C4DC seeks to address the inefficiencies of developing contractual agreements for public-private data collaboration by informing and guiding those seeking to establish a data collaborative by developing and making available a shared repository of relevant contractual clauses taken from existing legal agreements. Today TReNDS published “Partnerships Founded on Trust,” a brief capturing some initial findings from the C4DC initiative.

The Contractual Wheel of Data Collaboration [beta]

The Contractual Wheel of Data Collaboration [beta] — Stefaan G. Verhulst and Andrew Young, The GovLab

As part of the C4DC effort, and to support Data Stewards in the private sector and decision-makers in the public and civil sectors seeking to establish Data Collaboratives, The GovLab developed the Contractual Wheel of Data Collaboration [beta]. The Wheel seeks to capture key elements involved in data collaboration while demystifying contracts and moving beyond the type of legalese that can create confusion and barriers to experimentation.

The Wheel was developed based on an assessment of existing legal agreements, engagement with The GovLab-facilitated Data Stewards Network, and analysis of the key elements of our Data Collaboratives Methodology. It features 22 legal considerations organized across 6 operational categories that can act as a checklist for the development of a legal agreement between parties participating in a Data Collaborative:…(More)”.

How nudge theory is ageing well

Curated on Posted on by Stefaan Verhulst

Julian Baggini at the Financial Times: “A decade ago, Cass Sunstein and Richard Thaler’s book Nudge was on the desk of every serious politician and policy wonk. Its central thesis was alluringly simple: by changing the environment in which we make decisions — the “choice architecture” — people could be encouraged to do things that were good for them and for society without governments compelling them to do anything.

The idea hit the liberal sweet-spot, promising maximum social impact for minimal interference with personal freedom. In 2010, Britain’s government set up its Behavioural Insights Team — popularly known as the “nudge unit” — to put these ideas into practice.

Around the world, others followed. Sunstein is justly proud that 10m poor American children now get free breakfast and lunch during the academic year as a result of just one such intervention making enrolment for free school meals automatic.

Ten years on, Sunstein has produced two new books to win over the unconverted and boost the faith of true believers. One, On Freedom, is a tiny, commuter-friendly pamphlet between hard covers. The other, Trusting Nudges, co-authored with the behavioural economist Lucia A Reisch, is a short, thoughtful, measured and important analysis of what citizens actually think about nudging and why that matters — albeit with the dry, academic furniture of endless tables, footnotes and technical appendices.

Despite the stylistic gulf between them, the two books are best read together as a response to those who would like to give nudges the nudge, claiming that they are covert, manipulative, an insult to human agency and place too much trust in governments and too little on human reason. Not only that, but for all the hype, nudges only work at the margins, delivering relatively minor results without having any major impact on poverty, inequity or inequality.

On Freedom economically and elegantly takes apart the accusation that nudges undermine liberty. Sunstein rightly points out that a nudge is only a nudge by definition if it leaves the nudged able to choose otherwise. For example, the system adopted by several jurisdictions to put people on organ donation registers by default carries with it the right to opt out. Nor are the best nudges covert.

There may not be a sign at the canteen telling you that healthy foods have been put at the front because that’s where you’re more likely to choose them but organisations that adopt this as a policy can and should do so openly. Sunstein’s most important argument is that “we cannot wish choice architecture away”: something has to be on the supermarket shelves that people tend to take more from, something has to be the default for benefit claims. The question is not whether we nudge but how we do so: with forethought or without….(More)”

San Francisco teams up with Uber, location tracker on 911 call responses

Curated on Posted on by Stefaan Verhulst

Gwendolyn Wu at San Francisco Chronicle: “In an effort to shorten emergency response times in San Francisco, the city announced on Monday that it is now using location data from RapidSOS, a New York-based public safety tech company, and ride-hailing company Uber to improve location coordinates generated from 911 calls.

An increasing amount of emergency calls are made from cell phones, said Michelle Cahn, RapidSOS’s director of community engagement. The new technology should allow emergency responders to narrow down the location of such callers and replace existing 911 technology that was built for landlines and tied to home addresses.

Cell phone location data currently given to dispatchers when they receive a 911 call can be vague, especially if the person can’t articulate their exact location, according to the Department of Emergency Management.

But if a dispatcher can narrow down where the emergency is happening, that increases the chance of a timely response and better result, Cahn said.

“It doesn’t matter what’s going on with the emergency if we don’t know where it is,” she said.

RapidSOS shares its location data — collected by Apple and Google for their in-house map apps — free of charge to public safety agencies. San Francisco’s 911 call center adopted the data service in September 2018.

The Federal Communications Commission estimates agencies could save as many as 10,000 lives a year if they shave a minute off response times. Federal officials issued new rules to improve wireless 911 calls in 2015, asking mobile carriers to provide more accurate locations to call centers. Carriers are required to find a way to triangulate the caller’s location within 50 meters — a much smaller radius than the eight blocks city officials were initially presented in October when the caller dialed 911…(More)”.

Drones to deliver medicines to 12m people in Ghana

Curated on Posted on by Stefaan Verhulst

Neil Munshi in the Financial Times: “The world’s largest drone delivery network, ferrying 150 different medicines and vaccines, as well as blood, to 2,000 clinics in remote parts of Ghana, is set to be announced on Wednesday.

The network represents a big expansion for the Silicon Valley start-up Zipline, which began delivering blood in Rwanda in 2016 using pilotless, preprogrammed aircraft. The move, along with a new agreement in Rwanda signed in December, takes the company beyond simple blood distribution to more complicated vaccine and plasma deliveries.

“What this is going to show is that you can reach every GPS co-ordinate, you can serve everybody,” said Keller Rinaudo, Zipline chief executive. “Every human in that region or country [can be] within a 15-25 minute delivery of any essential medical product — it’s a different way of thinking about universal coverage.”

Zipline will deliver vaccines for yellow fever, polio, diptheria and tetanus which are provided by the World Health Organisation’s Expanded Project on Immunisation. The WHO will also use the company’s system for future mass immunisation programmes in Ghana.

Later this year, Zipline has plans to start operations in the US, in North Carolina, and in south-east Asia. The company said it will be able to serve 100m people within a year, up from the 22m that its projects in Ghana and Rwanda will cover.

In Ghana, Zipline said health workers will receive deliveries via a parachute drop within about 30 minutes of placing their orders by text message….(More)”.

Open data promotes citizen engagement at the local level

Curated on Posted on by Stefaan Verhulst

Afua Bruce at the Hill: “The city of Los Angeles recently released three free apps for its citizens: one to report broken street lighting, one to make 311 requests and one to get early alerts about earthquakes. Though it may seem like the city is just following a trend to modernize, the apps are part of a much larger effort to spread awareness of the more than 1,100 datasets that the city has publicized for citizens to view, analyze and share. In other words, the city has officially embraced the open data movement.

In the past few years, communities across the country have realized the power of data once only available to government. Often, the conversation about data focuses on criminal justice, because the demand for this data is being met by high-profile projects like Kamala Harris’ Open Justice Initiative, which makes California criminal justice data available to the citizenry and  the Open Data Policing Project, which provides a publicly searchable database of stop, search and use-of-force data. But the possibilities for data go far beyond justice and show the possibility for use in a variety of spaces, such as efforts to preserve local wildlife, track potholes and  understand community health trends….(More)”.

Selling civic engagement: A unique role for the private sector?

Curated on Posted on by Stefaan Verhulst

Rebecca Winthrop at Brookings: “Much has been written on the worrisome trends in Americans’ faith and participation in our nation’s democracy. According to the World Values Survey, almost 20 percent of millennials in the U.S. think that military rule or an authoritarian dictator is a “fairly good” form of government, and only 29 percent believe that living in a country that is governed democratically is “absolutely important.” In the last year, trust in American democratic institutions has dropped—only 53 percent of Americans view American democracy positively. This decline in faith and participation in our democracy has been ongoing for some time, as noted in the 2005 collection of essays, “Democracy At Risk: How Political Choices Undermine Citizen Participation, and What We Can Do About It.” The essays chart the “erosion of the activities and capacities of citizenship” from voting to broad civic engagement over the past several decades.

While civil society and government have been the actors most commonly addressing this worrisome trend, is there also a constructive role for the private sector to play? After all, compared to other options like military or authoritarian rule, a functioning democracy is much more likely to provide the conditions for free enterprise that business desires. One only has to look to the current events in Venezuela for a quick reminder of this.

Many companies do engage in a range of activities that broadly support civic engagement, from dedicating corporate social responsibility (CSR) dollars to civically-minded community activities to supporting employee volunteerism. These are worthy activities and should certainly continue, but given the crisis of faith in the foundations of our democratic process, the private sector could play a much bigger role in helping support a movement for renewed understanding of and participation in our political process. Many of the private sector’s most powerful tools for doing this lie not inside companies’ CSR portfolios but in their unique expertise in selling things. Every day companies leverage their expertise in influence—from branding to market-segmentation—to get Americans to use their products and services. What if this expertise were harnessed toward promoting civic understanding and engagement?

Companies could play a particularly useful role by tapping new resources to amplify existing good work and build increasing interest in civic engagement. Two ways of doing this could include the below….(More)”

Group decisions: When more information isn’t necessarily better

Curated on Posted on by Stefaan Verhulst

News Release from the Santa Fee Institute: “In nature, group decisions are often a matter of life or death. At first glance, the way certain groups of animals like minnows branch off into smaller sub-groups might seem counterproductive to their survival. After all, information about, say, where to find some tasty fish roe or which waters harbor more of their predators, would flow more freely and seem to benefit more minnows if the school of fish behaved as a whole. However, new research published in Philosophical Transactions of the Royal Society B sheds light on the complexity of collective decision-making and uncovers new insights into the benefits of the internal structure of animal groups.

In their paper, Albert Kao, a Baird Scholar and Omidyar Fellow at the Santa Fe Institute, and Iain Couzin, Director of the Max Planck Institute for Ornithology and Chair of Biodiversity and Collective Behavior at the University of Konstanz, simulate the information-sharing patterns of animals that prefer to interact with certain individuals over others. The authors’ modeling of such animal groups upends previously held assumptions about internal group structure and improves upon our understanding of the influence of group organization and environment on both the collective decision-making process and its accuracy.

Modular — or cliquey — group structure isolates the flow of communication between individuals, so that only certain animals are privy to certain pieces of information. “A feature of modular structure is that there’s always information loss,” says Kao, “but the effect of that information loss on accuracy depends on the environment.”

In simple environments, the impact of these modular groups is detrimental to accuracy, but when animals face many different sources of information, the effect is actually the opposite. “Surprisingly,” says Kao, “in complex environments, the information loss even helps accuracy in a lot of situations.” More information, in this case, is not necessarily better.

“Modular structure can have a profound — and unexpected — impact on the collective intelligence of groups,” says Couzin. “This may indeed be one of the reasons that we see internal structure in so many group-living species, from schooling fish and flocking birds to wild primate groups.”

Potentially, these new observations could be applied to many different kinds of social networks, from the migration patterns of birds to the navigation of social media landscapes to the organization of new companies, deepening our grasp of complex organization and collective behavior….(More)”.

(The paper, “Modular structure within groups causes information loss but can improve decision accuracy,” is part of a theme issue in the Philosophical Transactions of the Royal Society B entitled “Liquid Brains, Solid Brains: How distributed cognitive architectures process information.” The issue was inspired by a Santa Fe Institute working group and edited by Ricard Solé (Universitat Pompeu Fabra), Melanie Moses (University of New Mexico), and Stephanie Forrest (Arizona State University).

Finding Wisdom in Politically Polarized Crowds

Curated on Posted on by Stefaan Verhulst

Eamon Duede at Nature Research: “We were seeing that the consumption of ideas seemed deeply related io political alignment, and because our group (Knowledge Lab) is concerned with understanding the social dynamics involved in production of ideas, we began wondering whether and to what extent the political alignment of individuals contributes to a group’s ability to produce knowledge. A Wikipedia article is full of smuggled content and worked into a narrative by a diverse team of editors. Because those articles constitute knowledge, we were curious to know whether political polarization within those teams had an effect on the quality of that production. So, we decided to braid both strands of research together and look at the way in which individual political alignments and the polarization of the teams they form affect the quality of the work that is produced collaboratively on Wikipedia.

To answer this question, we turned not to the article itself, but the immense history of articles on Wikipedia. Every edit to every article, no matter how insignificant, is documented and saved in Wikipedia’s astonishingly massive archives. And every edit to every article, no matter how insignificant, is evaluated for its relevance or validity by the vast community of editors, both robotic and human. Remarkable teamwork has gone into producing the encyclopedia. Some people edit randomly, simply cleaning typos, adding citations, or contributing graffiti and vandalism (I’ve experimented with this, and it gets painted over very quickly, no matter where you put it). Yet, many people are genuinely purposeful in their work, and contribute specifically to topics on which they have both interest and knowledge. They tend and grow a handful of articles or a few broad topics like gardeners. We walked through the histories of these gardens, looking back at who made contributions here and there, how much they contributed, and where. We thought that editors who make frequent contributions to pages associated with American liberalism would hold left leaning opinions, and for conservatism opinions on the right. This was a controversial hypothesis, and many in the Wikipedia community felt that perhaps the opposite would be true, with liberals correcting conservative pages and conservatives kindly returning the favor -like weeding or applying pesticide. But a survey we conducted of active Wikipedia editors found that building a function over the relative number of bits they contributed to liberal versus conservative pages predicted more than a third of the probability that they identified as such and voted accordingly.

Following this validation, we assigned a political alignment score to hundreds of thousands of editors by looking at where they make contributions, and then examined the polarization within teams of editors that produced hundreds of thousands of Wikipedia articles in the broad topic areas of politics, social issues, and science. We found that when most members of a team have the same political alignment, whether conservative, liberal, or “independent”, the quality of the Wikipedia pages they produce is not as strong as those of teams with polarized compositions of editors (Shi et al. 2019).

The United States Senate is increasingly polarized, but largely balanced in its polarization. If the Senate was trying to write a Wikipedia article, would they produce a high quality article? If they are doing so on Wikipedia, following norms of civility and balance inscribed within Wikipedia’s policies and guidelines, committed to the production of knowledge rather than self-promotion, then the answer is probably “yes”. That is a surprising finding. We think that the reason for this is that the policies of Wikipedia work to suppress the kind of rhetoric and sophistry common in everyday discourse, not to mention toxic language and name calling. Wikipedia’s policies are intolerant of discussion that could distort balanced consideration of the edit and topic under consideration, and, given that these policies shut down discourse that could bias proposed edits, teams with polarized viewpoints have to spend significantly more time discussing and debating the content that is up for consideration for inclusion in an article. These diverse viewpoints seem to bring out points and arguments between team members that sharpen and refine the quality of the content they can collectively agree to. With assumptions and norms of respect and civility, political polarization can be powerful and generative….(More)”