Are individuals who are positive about AI applications also more unsure?

Worries about the use of artificial intelligence on recommending financial or medical advice, employee hiring, creating art, etc. get expressed through a mixture of someone’s default attitude towards the topic and an inherent unsureness. This “split” assumption makes possible the break-up of a recorded opinion instead of taking it at face value. Policy implications are immense. In this article, carried by Patterns https://doi.org/10.1016/j.patter.2025.101374, I show how, for some topics (such as offering financial advice), as people get surer, they get more positive, while for others (like AI’s impact on the job market, medical applications, etc.), sureness accompany negative-ness. Separation along demographic/political axes often correlate with separation of feeling-unsureness profiles. Uniform-binomial mixture densities help. As do Bentley University– Gallup Force For Good surveys. Free to read. hashtag#BentleyUResearch

Uniform-binomial mixture densities to gauge the impact of shifts in work operations on employees

As the ways we work evolve, it gets vital for employers to sense the impacts of tweaking work requirements on sections of their workforce. In this article: https://doi.org/10.1016/j.socimp.2025.100152 carried by Societal Impacts, I show how an opinion towards a proposed change (such as requiring checking emails outside of workdays, a return to a 5-day workweek, etc.) can be seen as a combination of a person’s default unsureness towards the topic, and their outlook (positive or negative) towards it. This decomposition can point out how seriously one needs to treat a recorded Likert-scale answer over key demographic dimensions.

“People who are sure of their outlook on returning to a five-day workweek generally feel the move would impact them negatively, but the tendency is stronger for women than for men. Certain industries where women are overrepresented may consider introducing additional benefits if this move is deemed necessary. In contrast, with limiting work-related matters outside of the workday people who are very sure of their opinion, feel the move would impact them positively (with men typically less unsure), again, the connection being more strong for women workers. Institutions that are considering introducing a “right to disconnect” policy may use this result to work out which employees – at what times in their career – require improved work-life balance or reduced burnout. Follow-up studies can be conducted on these individuals to test whether their productivity or trust in their employers, increased.”

Made with Bentley-Gallup Force for Good survey data. Free to Read. #BentleyUResearch

Cricket’s place in the Olympics: a look through “change-point” lenses

Quantifying excitement in sporting contexts has always been tricky and final scorelines prove to be demonstrably inadequate. In this article: https://doi.org/10.1177/17479541251372708 carried by The International Journal of Sports Science and Coaching Sage, I show how sudden shifts in continuous game-level features (like utility differentials in cricket or effective playing spaces in soccer) could supply a more realistic resolution, and bridge conceptual gaps between game categories. Through curve registration and functional analogues of unsupervised tasks (clustering, outlier detection), I argue how the game of cricket, by maintaining an optimum balance between chaos and predictability, merit a solid place at global sporting events like the Olympics (LA 28 and beyond). Free to read. #BentleyUResearch

A recipe for systems change: Predictive modeling and street-level bureaucracy among homeless services

Relocating homeless folks is a complex task involving human interactions and multiple layers of bureaucracy. Researchers have sensed that tweaking the ways of interacting could lead to increased success in terms of housing more people or finding homes at faster rates. Often, these require a loosening up of bureaucratic restrictions/conventions. Previously, researchers felt changes can be brought about in three broad areas of engagement: structural, relational, and transformative, with decreasing degrees of easiness. In this work, Curt Smith and I use data to confirm quantitatively that this is indeed so. We show that the pieces in each layer are not equal (as theoretical guesses seem to suggest): some changes are easier to bring about (for instance, relational) than others (for instance, power dynamics). We use conditional probabilities to install prediction possibilities. For instance, during a time-bound, oversight-free house-finding exercise, when changes in relationship, practice, and power dynamics (the easier ones) are sure, changes in mental structures (the harder ones) often follow with 50% chance. This is valid even if we know nothing about the state of the others (such as policy and resources), i.e., regardless of whether changes in those were or were not brought about. We also measure the impact of interruptions in workflow caused by preventable causes such as absenteeism. Interactive dashboards, free to read.

False identification in opinion surveys: Bayesian statistics to mellow out odd probability binaries

Parameter misidentification is common in statistics. Frequently, we end up estimating a quantity (like the potential impact a fresh policy will have on one individual) when something quite different was really sought after (like the amount of reach the policy will have on the full population). In an article: https://doi.org/10.1080/09332480.2025.2500900 I wrote with my research assistant, MSBA student Bahareh Zahirodini , we examine how Bayesian posteriors serve nicely here as frequentism falters. We use data from a recent Bentley University – Gallup survey, and examine how people feel on business practices around reducing carbon footprint, offering fair wages, creating a friendly workplace environment, etc. We offer non-crisp, non-binary probability envelopes: 54% chance, for instance, that at least 78% of the US population believes businesses ought to endorse DEI issues, test whether, and characterise how, thinking differs across demographic or political divides. Carried by Chance, American Statistical Association – ASA Taylor & Francis Research Insights Interactive dashboards, interesting visuals. Free to read.