Paper by Guy Grossman, Melina Platas and Jonathan Rodden: “We examine the effect on service delivery outcomes of a new information communication technology (ICT) platform that allows citizens to send free and anonymous messages to local government officials, thus reducing the cost and increasing the efficiency of communication about public services. In particular, we use a field experiment to assess the extent to which the introduction of this ICT platform improved monitoring by the district, effort by service providers, and inputs at service points in health, education and water in Arua District, Uganda. Despite relatively high levels of system uptake, enthusiasm of district officials, and anecdotal success stories, we find evidence of only marginal and uneven short-term improvements in health and water services, and no discernible long-term effects. Relatively few messages from citizens provided specific, actionable information about service provision within the purview and resource constraints of district officials, and users were often discouraged by officials’ responses. Our findings suggest that for crowd-sourced ICT programs to move from isolated success stories to long-term accountability enhancement, the quality and specific content of reports and responses provided by users and officials is centrally important….(More)”.
Artificial Intelligence and Public Policy
Paper by Adam D. Thierer, Andrea Castillo and Raymond Russell: “There is growing interest in the market potential of artificial intelligence (AI) technologies and applications as well as in the potential risks that these technologies might pose. As a result, questions are being raised about the legal and regulatory governance of AI, machine learning, “autonomous” systems, and related robotic and data technologies. Fearing concerns about labor market effects, social inequality, and even physical harm, some have called for precautionary regulations that could have the effect of limiting AI development and deployment. In this paper, we recommend a different policy framework for AI technologies. At this nascent stage of AI technology development, we think a better case can be made for prudence, patience, and a continuing embrace of “permissionless innovation” as it pertains to modern digital technologies. Unless a compelling case can be made that a new invention will bring serious harm to society, innovation should be allowed to continue unabated, and problems, if they develop at all, can be addressed later…(More)”.
Unnatural Surveillance: How Online Data Is Putting Species at Risk
Adam Welz at YaleEnvironment360: “…The burgeoning pools of digital data from electronic tags, online scientific publications, “citizen science” databases and the like – which have been an extraordinary boon to researchers and conservationists – can easily be misused by poachers and illegal collectors. Although a handful of scientists have recently raised concerns about it, the problem is so far poorly understood.
Today, researchers are surveilling everything from blue whales to honeybees with remote cameras and electronic tags. While this has had real benefits for conservation, some attempts to use real-time location data in order to harm animals have become known: Hunters have shared tips on how to use VHF radio signals from Yellowstone National Park wolves’ research collars to locate the animals. (Although many collared wolves that roamed outside the park have been killed, no hunter has actually been caught tracking tag signals.) In 2013, hackers in India apparently successfully accessed tiger satellite-tag data, but wildlife authorities quickly increased security and no tigers seem to have been harmed as a result. Western Australian government agents used a boat-mounted acoustic tag detector to hunt tagged white sharks in 2015. (At least one shark was killed, but it was not confirmed whether it was tagged). Canada’s Banff National Park last year banned VHF radio receivers after photographers were suspected of harassing tagged animals.
While there is no proof yet of a widespread problem, experts say it is often in researchers’ and equipment manufacturers’ interests to underreport abuse. Biologist Steven Cooke of Carleton University in Canada lead-authored a paper this year cautioning that the “failure to adopt more proactive thinking about the unintended consequences of electronic tagging could lead to malicious exploitation and disturbance of the very organisms researchers hope to understand and conserve.” The paper warned that non-scientists could easily buy tags and receivers to poach animals and disrupt scientific studies, noting that “although telemetry terrorism may seem far-fetched, some fringe groups and industry players may have incentives for doing so.”…(More)”.
Data-Driven Policy Making: The Policy Lab Approach
Societal challenges such as migration, poverty, and climate change can be considered ‘wicked problems’ for which no optimal solution exists. To address such problems, public administrations increasingly aim for data–driven policy making. Data-driven policy making aims to make optimal use of sensor data, and collaborate with citizens to co-create policy. However, few public administrations have realized this so far. Therefore, in this paper an approach for data-driven policy making is developed that can be used in the setting of a Policy Lab. A Policy Lab is an experimental environment in which stakeholders collaborate to develop and test policy. Based on literature, we first identify innovations in data-driven policy making. Subsequently, we map these innovations to the stages of the policy cycle. We found that most innovations are concerned with using new data sources in traditional statistics and that methodologies capturing the benefits of data-driven policy making are still under development. Further research should focus on policy experimentation while developing new methodologies for data-driven policy making at the same time….(More)”.
Algorithms in the Criminal Justice System: Assessing the Use of Risk Assessments in Sentencing
Priscilla Guo, Danielle Kehl, and Sam Kessler at Responsive Communities (Harvard): “In the summer of 2016, some unusual headlines began appearing in news outlets across the United States. “Secret Algorithms That Predict Future Criminals Get a Thumbs Up From the Wisconsin Supreme Court,” read one. Another declared: “There’s software used across the country to predict future criminals. And it’s biased against blacks.” These news stories (and others like them) drew attention to a previously obscure but fast-growing area in the field of criminal justice: the use of risk assessment software, powered by sophisticated and sometimes proprietary algorithms, to predict whether individual criminals are likely candidates for recidivism. In recent years, these programs have spread like wildfire throughout the American judicial system. They are now being used in a broad capacity, in areas ranging from pre-trial risk assessment to sentencing and probation hearings. This paper focuses on the latest—and perhaps most concerning—use of these risk assessment tools: their incorporation into the criminal sentencing process, a development which raises fundamental legal and ethical questions about fairness, accountability, and transparency. The goal is to provide an overview of these issues and offer a set of key considerations and questions for further research that can help local policymakers who are currently implementing or considering implementing similar systems. We start by putting this trend in context: the history of actuarial risk in the American legal system and the evolution of algorithmic risk assessments as the latest incarnation of a much broader trend. We go on to discuss how these tools are used in sentencing specifically and how that differs from other contexts like pre-trial risk assessment. We then delve into the legal and policy questions raised by the use of risk assessment software in sentencing decisions, including the potential for constitutional challenges under the Due Process and Equal Protection clauses of the Fourteenth Amendment. Finally, we summarize the challenges that these systems create for law and policymakers in the United States, and outline a series of possible best practices to ensure that these systems are deployed in a manner that promotes fairness, transparency, and accountability in the criminal justice system….(More)”.
Who Falls for Fake News? The Roles of Analytic Thinking, Motivated Reasoning, Political Ideology, and Bullshit Receptivity
Paper by Gordon Pennycook and David G. Rand: “Inaccurate beliefs pose a threat to democracy and fake news represents a particularly egregious and direct avenue by which inaccurate beliefs have been propagated via social media. Here we investigate the cognitive psychological profile of individuals who fall prey to fake news. We find a consistent positive correlation between the propensity to think analytically – as measured by the Cognitive Reflection Test (CRT) – and the ability to differentiate fake news from real news (“media truth discernment”). This was true regardless of whether the article’s source was indicated (which, surprisingly, also had no main effect on accuracy judgments). Contrary to the motivated reasoning account, CRT was just as positively correlated with media truth discernment, if not more so, for headlines that aligned with individuals’ political ideology relative to those that were politically discordant. The link between analytic thinking and media truth discernment was driven both by a negative correlation between CRT and perceptions of fake news accuracy (particularly among Hillary Clinton supporters), and a positive correlation between CRT and perceptions of real news accuracy (particularly among Donald Trump supporters). This suggests that factors that undermine the legitimacy of traditional news media may exacerbate the problem of inaccurate political beliefs among Trump supporters, who engaged in less analytic thinking and were overall less able to discern fake from real news (regardless of the news’ political valence). We also found consistent evidence that pseudo-profound bullshit receptivity negatively correlates with perceptions of fake news accuracy; a correlation that is mediated by analytic thinking. Finally, analytic thinking was associated with an unwillingness to share both fake and real news on social media. Our results indicate that the propensity to think analytically plays an important role in the recognition of misinformation, regardless of political valence – a finding that opens up potential avenues for fighting fake news….(More)”.
What does it mean to be differentially private?
Paul Francis at IAPP: “Back in June 2016, Apple announced it will use differential privacy to protect individual privacy for certain data that it collects. Though already a hot research topic for over a decade, this announcement introduced differential privacy to the broader public. Before that announcement, Google had already been using differential privacy for collecting Chrome usage statistics. And within the last month, Uber announced that they too are using differential privacy.
If you’ve done a little homework on differential privacy, you may have learned that it provides provable guarantees of privacy and concluded that a database that is differentially private is, well, private — in other words, that it protects individual privacy. But that isn’t necessarily the case. When someone says, “a database is differentially private,” they don’t mean that the database is private. Rather, they mean, “the privacy of the database can be measured.”
Really, it is like saying that “a bridge is weight limited.” If you know the weight limit of a bridge, then yes, you can use the bridge safely. But the bridge isn’t safe under all conditions. You can exceed the weight limit and hurt yourself.
The weight limit of bridges is expressed in tons, kilograms or number of people. Simplifying here a bit, the amount of privacy afforded by a differentially private database is expressed as a number, by convention labeled ε (epsilon). Lower ε means more private.
All bridges have a weight limit. Everybody knows this, so it sounds dumb to say, “a bridge is weight limited.” And guess what? All databases are differentially private. Or, more precisely, all databases have an ε. A database with no privacy protections at all has an ε of infinity. It is pretty misleading to call such a database differentially private, but mathematically speaking, it is not incorrect to do so. A database that can’t be queried at all has an ε of zero. Private, but useless.
In their paper on differential privacy for statistics, Cynthia Dwork and Adam Smith write, “The choice of ε is essentially a social question. We tend to think of ε as, say, 0.01, 0.1, or in some cases, ln 2 or ln 3.” The natural logarithm of 3 (ln 3) is around 1.1….(More)”.
Crowdsourcing citizen science: exploring the tensions between paid professionals and users
Jamie Woodcock et al in the Journal of Peer Production: “This paper explores the relationship between paid labour and unpaid users within the Zooniverse, a crowdsourced citizen science platform. The platform brings together a crowd of users to categorise data for use in scientific projects. It was initially established by a small group of academics for a single astronomy project, but has now grown into a multi-project platform that has engaged over 1.3 million users so far. The growth has introduced different dynamics to the platform as it has incorporated a greater number of scientists, developers, links with organisations, and funding arrangements—each bringing additional pressures and complications. The relationships between paid/professional and unpaid/citizen labour have become increasingly complicated with the rapid expansion of the Zooniverse. The paper draws on empirical data from an ongoing research project that has access to both users and paid professionals on the platform. There is the potential through growing peer-to-peer capacity that the boundaries between professional and citizen scientists can become significantly blurred. The findings of the paper, therefore, address important questions about the combinations of paid and unpaid labour, the involvement of a crowd in citizen science, and the contradictions this entails for an online platform. These are considered specifically from the viewpoint of the users and, therefore, form a new contribution to the theoretical understanding of crowdsourcing in practice….(More)”.
Artificial Intelligence for Citizen Services and Government
Paper by Hila Mehr: “From online services like Netflix and Facebook, to chatbots on our phones and in our homes like Siri and Alexa, we are beginning to interact with artificial intelligence (AI) on a near daily basis. AI is the programming or training of a computer to do tasks typically reserved for human intelligence, whether it is recommending which movie to watch next or answering technical questions. Soon, AI will permeate the ways we interact with our government, too. From small cities in the US to countries like Japan, government agencies are looking to AI to improve citizen services.
While the potential future use cases of AI in government remain bounded by government resources and the limits of both human creativity and trust in government, the most obvious and immediately beneficial opportunities are those where AI can reduce administrative burdens, help resolve resource allocation problems, and take on significantly complex tasks. Many AI case studies in citizen services today fall into five categories: answering questions, filling out and searching documents, routing requests, translation, and drafting documents. These applications could make government work more efficient while freeing up time for employees to build better relationships with citizens. With citizen satisfaction with digital government offerings leaving much to be desired, AI may be one way to bridge the gap while improving citizen engagement and service delivery.
Despite the clear opportunities, AI will not solve systemic problems in government, and could potentially exacerbate issues around service delivery, privacy, and ethics if not implemented thoughtfully and strategically. Agencies interested in implementing AI can learn from previous government transformation efforts, as well as private-sector implementation of AI. Government offices should consider these six strategies for applying AI to their work: make AI a part of a goals-based, citizen-centric program; get citizen input; build upon existing resources; be data-prepared and tread carefully with privacy; mitigate ethical risks and avoid AI decision making; and, augment employees, do not replace them.
This paper explores the various types of AI applications, and current and future uses of AI in government delivery of citizen services, with a focus on citizen inquiries and information. It also offers strategies for governments as they consider implementing AI….(More)”
Journal tries crowdsourcing peer reviews, sees excellent results
Chris Lee at ArsTechnica: “Peer review is supposed to act as a sanity check on science. A few learned scientists take a look at your work, and if it withstands their objective and entirely neutral scrutiny, a journal will happily publish your work. As those links indicate, however, there are some issues with peer review as it is currently practiced. Recently, Benjamin List, a researcher and journal editor in Germany, and his graduate assistant, Denis Höfler, have come up with a genius idea for improving matters: something called selected crowd-sourced peer review….
My central point: peer review is burdensome and sometimes barely functional. So how do we improve it? The main way is to experiment with different approaches to the reviewing process, which many journals have tried, albeit with limited success. Post-publication peer review, when scientists look over papers after they’ve been published, is also an option but depends on community engagement.
But if your paper is uninteresting, no one will comment on it after it is published. Pre-publication peer review is the only moment where we can be certain that someone will read the paper.
So, List (an editor for Synlett) and Höfler recruited 100 referees. For their trial, a forum-style commenting system was set up that allowed referees to comment anonymously on submitted papers but also on each other’s comments as well. To provide a comparison, the papers that went through this process also went through the traditional peer review process. The authors and editors compared comments and (subjectively) evaluated the pros and cons. The 100-person crowd of researchers was deemed the more effective of the two.
The editors found that it took a bit more time to read and collate all the comments into a reviewers’ report. But it was still faster, which the authors loved. Typically, it took the crowd just a few days to complete their review, which compares very nicely to the usual four to six weeks of the traditional route (I’ve had papers languish for six months in peer review). And, perhaps most important, the responses were more substantive and useful compared to the typical two-to-four-person review.
So far, List has not published the trial results formally. Despite that, Synlett is moving to the new system for all its papers.
Why does crowdsourcing work?
Here we get back to something more editorial. I’d suggest that there is a physical analog to traditional peer review, called noise. Noise is not just a constant background that must be overcome. Noise is also generated by the very process that creates a signal. The difference is how the amplitude of noise grows compared to the amplitude of signal. For very low-amplitude signals, all you measure is noise, while for very high-intensity signals, the noise is vanishingly small compared to the signal, even though it’s huge compared to the noise of the low-amplitude signal.
Our esteemed peers, I would argue, are somewhat random in their response, but weighted toward objectivity. Using this inappropriate physics model, a review conducted by four reviewers can be expected (on average) to contain two responses that are, basically, noise. By contrast, a review by 100 reviewers may only have 10 responses that are noise. Overall, a substantial improvement. So, adding the responses of a large number of peers together should produce a better picture of a scientific paper’s strengths and weaknesses.
Didn’t I just say that reviewers are overloaded? Doesn’t it seem that this will make the problem worse?
Well, no, as it turns out. When this approach was tested (with consent) on papers submitted to Synlett, it was discovered that review times went way down—from weeks to days. And authors reported getting more useful comments from their reviewers….(More)”.