Searcharxiv⌕ Search

arXiv subjects

Jason Alexander

Publications and source records attributed to Jason Alexander.

4 recordsLinked to original sources

A Crowd-Audited Atlas of Implicit Bias in LLM-Powered Agents

Prior work has shown that large language models (LLMs) express implicit biases in high-publicity, one-off decision scenarios such as hiring and intelligence assessment. However, as people increasingly employ autonomous, LLM-powered AI agents that make decisions on their behalf in doing day-to-day tasks, it is of critical importance that we understand how implicit biases, based on sociodemographic assumptions, emerge across a broad milieu of workaday decision scenarios. We used the 17 domains of daily activity in the American Time Use Survey (ATUS) to generate an atlas of 506 decision scenarios in which LLMs --- endowed with personas representing different races, gender identities, and political ideologies --- expressed significant implicit bias. Next, in a crowdsourced audit, 500 participants made 8,010 binary choices about which implicitly biased decision scenarios most deserved further review. We found that participants prioritized decision scenarios that had broad relevance (e.g., about food choices), were high stakes (e.g., about health), and where all members of a simulated sociodemographic group made unanimous decisions (because it was a clear indication of stereotyping). They were less swayed by the magnitude of an observed disparity or the specific ATUS domain the decision scenario represented. When examining the rationale behind their prioritization choices, we found that participants often interpreted disparities as traits of demographic groups and judged whether those traits seemed plausible, often based on their own internalized stereotypes. We discuss implications for building personalized AI agents, digital twins, social simulations, and for using crowdsourced bias audits.

cs.CY↗

The Frequency and Mass-Ratio Distribution of Binaries in Clusters -- III: Probabilistic Generative Modelling of Six Young Open Clusters

We apply probabilistic generative modelling of colour-magnitude diagrams to six young Galactic open star clusters and determine their mass functions, binary mass-ratio distributions, and the frequencies of binary stars. We find that younger clusters tend to exhibit a higher incidence of binaries than their older counterparts. The mass-ratio distribution is fairly flat for the clusters with one exception that exhibits a sharp increase for $q\gtrsim0.9$. The ratio of the number of cluster binaries for which $q>0.75$ to the number of binaries for which $q>0.5$ (referred to as $FQ_{75}$) ranges from $\sim0.4 - 0.8$. This metric increases with the binary-star frequency of a cluster, but declines with cluster age. This may be due to non-ionizing 3-body dynamical processing of a primordial population of close binaries with initial mass ratios, $q \simeq 1$.

astro-ph.GA↗

Can GPT-4 Models Detect Misleading Visualizations?

The proliferation of misleading visualizations online, particularly during critical events like public health crises and elections, poses a significant risk. This study investigates the capability of GPT-4 models (4V, 4o, and 4o mini) to detect misleading visualizations. Utilizing a dataset of tweet-visualization pairs containing various visual misleaders, we test these models under four experimental conditions with different levels of guidance. We show that GPT-4 models can detect misleading visualizations with moderate accuracy without prior training (naive zero-shot) and that performance notably improves when provided with definitions of misleaders (guided zero-shot). However, a single prompt engineering technique does not yield the best results for all misleader types. Specifically, providing the models with misleader definitions and examples (guided few-shot) proves more effective for reasoning misleaders, while guided zero-shot performs better for design misleaders. This study underscores the feasibility of using large vision-language models to detect visual misinformation and the importance of prompt engineering for optimized detection accuracy.

cs.CV↗

Bose-Einstein condensation transition studies for atoms confined in Laguerre-Gaussian laser modes

Multiply-connected traps for cold, neutral atoms fix vortex cores of quantum gases. Laguerre-Gaussian laser modes are ideal for such traps due to their phase stability. We report theoretical calculations of the Bose-Einstein condensation transition properties and thermal characteristics of neutral atoms trapped in multiply connected geometries formed by Laguerre-Gaussian LG{p}{l} beams. Specifically, we consider atoms confined to the anti-node of a LG{0}{1} laser mode detuned to the red of an atomic resonance frequency, and those confined in the node of a blue-detuned LG{1}{1} beam. We compare the results of using the full potential to those approximating the potential minimum with a simple harmonic oscillator potential. We find that deviations between calculations of the full potential and the simple harmonic oscillator can be up to 3%-8% for trap parameters consistent with typical experiments.

physics.atom-ph↗