When Do We Trust AI’s Recommendations More Than People’s?

Curated on Posted on by Stefaan Verhulst

Chiara Longoni and Luca Cian at Harvard Business School: “More and more companies are leveraging technological advances in machine learning, natural language processing, and other forms of artificial intelligence to provide relevant and instant recommendations to consumers. From Amazon to Netflix to REX Real Estate, firms are using AI recommenders to enhance the customer experience. AI recommenders are also increasingly used in the public sector to guide people to essential services. For example, the New York City Department of Social Services uses AI to give citizens recommendations on disability benefits, food assistance, and health insurance.

However, simply offering AI assistance won’t necessarily lead to more successful transactions. In fact, there are cases when AI’s suggestions and recommendations are helpful and cases when they might be detrimental. When do consumers trust the word of a machine, and when do they resist it? Our research suggests that the key factor is whether consumers are focused on the functional and practical aspects of a product (its utilitarian value) or focused on the experiential and sensory aspects of a product (its hedonic value).

In an article in the Journal of Marketing — based on data from over 3,000 people who took part in 10 experiments — we provide evidence supporting for what we call a word-of-machine effect: the circumstances in which people prefer AI recommenders to human ones.

The word-of-machine effect.

The word-of-machine effect stems from a widespread belief that AI systems are more competent than humans in dispensing advice when utilitarian qualities are desired and are less competent when the hedonic qualities are desired. Importantly, the word-of-machine effect is based on a lay belief that does not necessarily correspond to the reality. The fact of the matter is humans are not necessarily less competent than AI at assessing and evaluating utilitarian attributes. Vice versa, AI is not necessarily less competent than humans at assessing and evaluating hedonic attributes….(More)”.

To Fight Polarization, Ask, “How Does That Policy Work?”

Curated on Posted on by Stefaan Verhulst

Article by Alex Chesterfield and Kate Coombs: “…One reason for this effect, and for the polarizing outcome, is we often overestimate our understanding of how political policies work. In this case, the more omniscient we think we are, the easier it is to ignore alternative facts or ideas. This phenomenon has a name—the illusion of explanatory depth (IOED). Unless explicitly tested, individuals can remain largely unaware of the shallowness of their own understanding of the things they think they understand—such as the mechanics of a bicycle, or how the policy they support or despise will actually work.

Researchers have started to explore what happens to political attitudes when you explicitly test people on how much they actually know about a policy. When people discover that they don’t know as much as they thought they did, something interesting happens: their political attitudes become less extreme….

Some countries and institutions are already using these insights to improve decision-making on divisive topics. Deliberative democracy, which plays out in the form of citizens’ assemblies and juries, where a small group of people (12-24) come together to deliberate on an issue, provide time and information to encourage participants to generate explanations—rather than justifications based on values, hearsay, or feelings—for their positions. Participants also tend to be representative of the general population; research suggests that increasing contact between diverse individuals could also help diminish affective polarization by shrinking the prejudices we form when making assumptions about the “other” that are based on reductive stereotypes, rather than real, complex people.

Outside of juries and citizens assemblies, countries like Ireland have used deliberative democracy to address a range of complex and highly polarized issues including same-sex marriage, access to abortion, and climate change. U.K. politicians from both sides of the aisle have called for a Brexit assembly to try and break the U.K. political deadlock. Will it work? We don’t know yet, and we’d encourage researchers to continue to study this topic. In the meantime, we can each begin by  confronting our own ignorance. Before committing to a position or policy, ask yourself to explain mechanistically how you think it will bring about the intended outcome. Do you really understand it?

Test your own mechanistic reasoning. Pick a topic you feel strongly about: climate change, Brexit, Immigration, gun laws, assisted suicide/legal euthanasia. Instead of justifying why you support a particular position so strongly, try to explain how it might lead to a particular outcome….(More)”

Why Doubt Is Essential to Science

Curated on Posted on by Stefaan Verhulst

Liv Grjebine at Scientific American: “The confidence people place in science is frequently based not on what it really is, but on what people would like it to be. When I asked students at the beginning of the year how they would define science, many of them replied that it is an objective way of discovering certainties about the world. But science cannot provide certainties. For example, a majority of Americans trust science as long as it does not challenge their existing beliefs. To the question “When science disagrees with the teachings of your religion, which one do you believe?,” 58 percent of North Americans favor religion; 33 percent science; and 6 percent say “it depends.”

But doubt in science is a feature, not a bug. Indeed, the paradox is that science, when properly functioning, questions accepted facts and yields both new knowledge and new questions—not certainty. Doubt does not create trust, nor does it help public understanding. So why should people trust a process that seems to require a troublesome state of uncertainty without always providing solid solutions?


As a historian of science, I would argue that it’s the responsibility of scientists and historians of science to show that the real power of science lies precisely in what is often perceived as its weakness: its drive to question and challenge a hypothesis. Indeed, the scientific approach requires changing our understanding of the natural world whenever new evidence emerges from either experimentation or observation. Scientific findings are hypotheses that encompass the state of knowledge at a given moment. In the long run, many of are challenged and even overturned. Doubt might be troubling, but it impels us towards a better understanding; certainties, as reassuring as they may seem, in fact undermine the scientific process….(More)”.

Science as Scorekeeping

Curated on Posted on by Stefaan Verhulst


Brendan Foht at New Atlantis: “If there is one thing about the coronavirus pandemic that both sides of the political spectrum seem to agree on, it’s that the science that bears on it should never be “politicized.” From the left, former CDC directors of the Obama and Clinton administrations warn of how the Trump administration has politicized the agency’s science: “The only valid reason to change released guidelines is new information and new science — not politics.” From the right, the Wall Street Journal frets about the scientific journal Nature publishing a politically charged editorial about why China shouldn’t be blamed for the coronavirus: “Political pressure has distorted scientific judgment.” What both sides assume is that political authorities should defer to scientists on important decisions about the pandemic, but only insofar as science itself is somehow kept free from politics.

But politicization, and even polarization, are not always bad for science. There is much about how we can use science to respond to the pandemic that is inescapably political, and that we cannot simply leave to scientists to decide.

There is, however, a real problem with how political institutions in the United States have engaged with science. Too much of the debate over coronavirus science has centered on how bad the disease really is, with the administration downplaying its risks and the opposition insisting on its danger. One side sees the scientists warning of peril as a political obstacle that must be overcome. The other side sees them as authorities to whom we must defer, not as servants of the public who could be directed toward solving the problem. The false choice between these two perspectives on how science relates to politics obscures a wide range of political choices the country faces about how we can make use of our scientific resources in responding to the pandemic….(More)”.

If data is 21st century oil, could foundations be the right owners?

Curated on Posted on by Stefaan Verhulst

Felix Oldenburg at Alliance: “What are the best investments for a foundation? This important question is one many foundation professionals are revisiting in light of low interest rates, high market volatility, and fears of deep economic trouble ahead. While stories of success certainly exist and are worth learning from, even the notorious lack of data cannot obscure the inconvenient truth that the idea of traditional endowments is in trouble.

I would argue that in order to unleash the potential of foundations, we should turn the question around, perhaps back on its feet: For which assets are foundations the best owners?

In the still dawning digital age, one fascinating answer may stare you right in the face as you read this. How much is your personal data worth? Your social media information, search and purchase history, they are the source of much of the market value of the fastest growing sector of our time. A rough estimate of market valuation of the major social platforms divided by their active users arrives at more than $1,000 USD per user, not differentiating by location or other factors. This sum is more than the median per capita wealth in about half the world’s countries. And if the trend continues, this value may continue to grow – and with it the big question of how to put one of the most valuable resource of our time to use for the good of all.

Acting as guardians of digital commons, data-endowed foundations could negotiate conditions for the commercial use of its assets, and invest the income to create equal digital opportunities, power 21st century education, and fight climate change.

Foundation ownership in the data sector may sound like a wild idea at first. Yet foundations and their predecessors have played the role of purpose-driven owners of critical assets and infrastructures throughout history. Monasteries (called ‘Stifte’ in German, the root of the German word for foundations) have protected knowledge and education in libraries, and secured health care in hospitals. Trusts have created affordable much of the social housing in the exploding cities of the 19th century. The German Marshall Plan created an endowment for economic recovery that is still in existence today.

The proposition is simple: Independent ownership for the good of all, beyond the commercial or national interests of individual corporations of governments, in perpetuity. Acting as guardians of digital commons, data-endowed foundations could negotiate conditions for the commercial use of its assets, and invest the income to create equal digital opportunities, power 21st century education, and fight climate change. An ideal model of ownership would also include a form of governance exercised by the users themselves through digital participation and elections. A foundation really only relies on one thing, a stable frame of rights in its legal home country. This is far from a trivial condition, but again history shows how many foundations have survived depressions, wars, and revolutions….(More)”

COVID-19 Is Challenging Medical and Scientific Publishing

Curated on Posted on by Stefaan Verhulst

Article by By Vilas Dhar, Amy Brand & Stefano Bertozzi: “We need a transformation in how early data is shared. But the urgent need for peer-reviewed science, coupled with the potential harms of unreviewed publication, has set the stage for a public discussion on the future of academic publishing. It’s clear that we need rapid, transparent peer review that allows reviewers, authors, and readers to engage with one another, and for dynamic use of technology to accelerate publishing timelines without reducing academic rigor or researcher accountability. However, the field of academic publishing will need significant financial support to catalyze these changes.

Philanthropic organizations, as longtime supporters of scientific research, must be at the vanguard of the effort to fund improvements in how science is curated, reviewed, and published. When the MIT Press first began to address the need for the rapid dissemination of COVID-19-related research and scholarship—by making a selection relevant e-books and journal articles freely available, as well as developing a new, rapid publication model for books, under the imprint First Reads—senior staff were interested in undertaking bolder efforts to address the specific problems engendered by the pandemic. The proliferation of preprints related to COVID-19 was already apparent, as was the danger of un-vetted science seeding mainstream media stories with deleterious results.

Rapid Reviews: COVID-19 (RR:C19) is an innovation in open publishing that allows for rigorous, transparent peer review that is publicly shared in advance of publication. We believe that pushing the peer review process further upstream—so that it occurs at the preprint stage—will benefit a wide variety of stakeholders: journalists, clinicians, researchers, and the public at large.  …

With this and future efforts, we’ve identified five key opportunities to align academic publishing priorities with the public good:

  1. Transparency: Redesign and incentivize the peer review process to publish all peer reviews alongside primary research, reducing duplicate reviews and allowing readers and authors to understand and engage with the critiques.
  2. Accountability: The roles of various authors on any given manuscript should be clearly defined and presented for the readers. When datasets are used, one or more of the authors should have explicit responsibility for verifying the integrity of the data and should document that verification process within the paper’s methodology section.
  3. Urgency: Scientific research can be slow moving and time consuming. Publishing data does not have to be. Publishing houses should build networks of experts who are able to dedicate time to scrutinizing papers in a timely manner with the goal of rapid review with rigor.
  4. Digital-First Publishing: While science is a dynamic process of continued learning and exploration, much of scientific publishing conforms to outdated print models. Academic journals should explore opportunities to deploy AI-powered tools to identify peer-reviewers or preprint scholarship and digital publishing platforms to enable more visible communication and collaboration about research findings. Not only can reviews be closer to real-time, but authors can easily respond and modify their work for continuous quality improvement.
  5. Funding: Pioneering new solutions in academic publishing will require significant trial and error, at a time when traditional business models such as library subscriptions are in decline. Philanthropies should step forward to provide catalytic risk financing, testing new models and driving social good outcomes….(More)”.

UK passport photo checker shows bias against dark-skinned women

Curated on Posted on by Stefaan Verhulst

Maryam Ahmed at BBC News: “Women with darker skin are more than twice as likely to be told their photos fail UK passport rules when they submit them online than lighter-skinned men, according to a BBC investigation.

One black student said she was wrongly told her mouth looked open each time she uploaded five different photos to the government website.

This shows how “systemic racism” can spread, Elaine Owusu said.

The Home Office said the tool helped users get their passports more quickly.

“The indicative check [helps] our customers to submit a photo that is right the first time,” said a spokeswoman.

“Over nine million people have used this service and our systems are improving.

“We will continue to develop and evaluate our systems with the objective of making applying for a passport as simple as possible for all.”

Skin colour

The passport application website uses an automated check to detect poor quality photos which do not meet Home Office rules. These include having a neutral expression, a closed mouth and looking straight at the camera.

BBC research found this check to be less accurate on darker-skinned people.

More than 1,000 photographs of politicians from across the world were fed into the online checker.

The results indicated:

  • Dark-skinned women are told their photos are poor quality 22% of the time, while the figure for light-skinned women is 14%
  • Dark-skinned men are told their photos are poor quality 15% of the time, while the figure for light-skinned men is 9%

Photos of women with the darkest skin were four times more likely to be graded poor quality, than women with the lightest skin….(More)”.

How Not to Kill People With Spreadsheets

Curated on Posted on by Stefaan Verhulst

David Gerard at Foreign Policy: “The U.K.’s response to COVID-19 is widely regarded as scattershot and haphazard. So how did they get here?

Excel is a top-of-the-line spreadsheet tool. A spreadsheet is good for quickly modeling a problem—but too often, organizations cut corners and press the cardboard-and-string mock-up into production, instead of building a robust and unique system based on the Excel proof of concept.

Excel is almost universally misused for complex data processing, as in this case—because it’s already present on your work computer and you don’t have to spend months procuring new software. So almost every business has at least one critical process that relies on a years-old spreadsheet set up by past staff members that nobody left at the company understands.

That’s how the U.K. went wrong. An automated process at Public Health England (PHE) transformed the incoming private laboratory test data (which was in text-based CSV files) into Excel-format files, to pass to the Serco Test and Trace teams’ dashboards.

Unfortunately, the process produced XLS files—an outdated Excel format that went extinct in 2003—which had a limit of 65,536 rows, rather than the around 1 million-row limit in the more recent XLSX format. With several lines of data per patient, this meant a sheet could only hold 1,400 cases. Further cases just fell off the end.

Technicians at PHE monitoring the dashboards noticed on Oct. 2 that not all data that had been sent in was making it out the other end. The data was corrected the next day, and PHE announced the issue the day after.

It’s not clear if the software at PHE was an Excel spreadsheet or an in-house program using the XLS format for data interchange—the latter would explain why PHE stated that replacing it might take months—but the XLS format would have been used on the assumption that Excel was universal.

And even then, a system based on Excel-format files would have been an improvement over earlier systems—the system for keeping a count of COVID-19 cases in the U.K. was, as of May, still based on data handwritten on cards….

The process that went wrong was a workaround for a contract issue: The government’s contract with Deloitte to run the testing explicitly stipulated that the company did not have to report “Pillar 2” (general public testing) positive cases to PHE at all.

Since a test-and-trace system is not possible without this data, PHE set up feeds for the data anyway, as CSV text files directly from the testing labs. The data was then put into this system—the single system that serves as the bridge between testing and tracing, for all of England. PHE had to put in place technological duct tape to make a system of life-or-death importance work at all….

The Brookings Institution report Doomed: Challenges and solutions to government IT projects lists factors to consider when outsourcing government information technology. The outsourcing of tracking and tracing is an example where the government has assumed all of the risk, and the contractor assumes all of the profit. PHE did one thing that you should never do: It outsourced a core function. Running a call center or the office canteen? You can outsource it. Tracing a pandemic? You must run it in-house.

If you need outside expertise for a core function, use contractors working within a department. Competing with the private sector on pay can be an issue, but a meaningful project can be a powerful incentive….(More)”.

Scotland’s future vision discussed today in first Citizens’ Assembly

Curated on Posted on by Stefaan Verhulst

Article by Richard Mason: “The group of 100 broadly representative Scots have been meeting throughout the year to discuss some of the country’s major constitutional issues.

Members have been asked to consider three questions, the first of which is: “What kind of country are we seeking to build?”

The assembly will meet online to develop the vision, having examined issues such as finances and taxation, and discussed how decisions are taken for and about Scotland. A report of the meeting will be published on October 9

The other two parts of the Assembly’s remit – how to best overcome the challenges the country faces, including Brexit, and how to empower people to make “informed choices” about Scotland’s future – will be addressed in a final report by the end of the year.

Assembly convener Kate Wimpress said: “The meeting this weekend will see a group of people from all walks of life across Scotland come together to agree a shared vision of our country’s future.

“The Citizens’ Assembly’s vision for Scotland will help give a roadmap for the country at an uncertain and difficult time.

“Our members have worked hard together across the months, and it’s exciting to witness their efforts now coming to fruition.”

First Minister Nicola Sturgeon announced the creation of the Citizens’ Assembly and outlined its remit, but she stressed it would be independent from Government following criticism it was set up to garner independence support.

Constitution Secretary Michael Russell said the Scottish Government is spending £1.37 million to fund six assembly meetings, which were held in person before moving online following the coronavirus lockdown….(More)”

Social license for the use of big data in the COVID-19 era

Curated on Posted on by Stefaan Verhulst

Commentary by James A. Shaw, Nayha Sethi & Christine K. Cassel: “… Social license refers to the informal permissions granted to institutions such as governments or corporations by members of the public to carry out a particular set of activities. Much of the literature on the topic of social license has arisen in the field of natural resources management, emphasizing issues that include but go beyond environmental stewardship4. In their seminal work on social license in the pulp and paper industry, Gunningham et al. defined social license as the “demands and expectations” placed on organizations by members of civil society which “may be tougher than those imposed by regulation”; these expectations thereby demand actions that go beyond existing legal rules to demonstrate concern for the interests of publics. We use the plural term “publics” as opposed to the singular “public” to illustrate that stakeholder groups to which organizations must appeal are often diverse and varied in their assessments of whether a given organizational activity is acceptable6. Despite the potentially fragmented views of various publics, the concept of social license is considered in a holistic way (either an organization has it or does not). Social license is closely related to public trust, and where publics view a particular institution as trustworthy it is more likely to have social license to engage in activities such as the collection and use of personal data7.

The question of how the leaders of an organization might better understand whether they have social license for a particular set of activities has also been addressed in the literature. In a review of literature on social license, Moffat et al. highlighted disagreement in the research community about whether social license can be accurately measured4. Certain groups of researchers emphasize that because of the intangible nature of social license, accurate measurement will never truly be possible. Others propose conceptual models of the determinants of social license, and establish surveys that assess those determinants to indicate the presence or absence of social license in a given context. However, accurate measurement of social license remains a point of debate….(More)”.