Browse Articles

Various methods exist to mitigate bias in AI models, including "Constitutional AI," a technique which guides the AI to behave according to a list of rules and principles. Lo, Poosarla, Singhal, Li, Fu, and Mui investigate whether constitutional AI can reduce bias in AI outputs on political topics.

The authors looked at the relation between stress and motivation in high school students with 4+ AP classes that also played a varsity sport. No distinct correlation was observed, however, results indicated that there are other factors at play that influence both stress and motivation.

The authors develop a new method for training machine learning algorithms to differentiate between hate speech and cultural speech in online platforms.

Using facial recognition as a use-case scenario, we attempt to identify sources of bias in a model developed using transfer learning. To achieve this task, we developed a model based on a pre-trained facial recognition model, and scrutinized the accuracy of the model’s image classification against factors such as age, gender, and race to observe whether or not the model performed better on some demographic groups than others. By identifying the bias and finding potential sources of bias, his work contributes a unique technical perspective from the view of a small scale developer to emerging discussions of accountability and transparency in AI.

Voice pitch affects perceived authoritativeness, competency, and leadership capacity. In this study, the authors suggest that examining certain measures of brain activity collected using an affordable EEG could predict advertising effectiveness, which may be invaluable in future neuromarketing research. Understanding voice pitch and other factors that cause implicit bias may allow significant advances in marketing, facilitating business success.

Malaise traps are commonly used to collect flying insects for a variety of research. In this study, researchers hypothesized the attractants used in these traps may create bias in insect studies that could lead to misinterpreted data. To test this hypothesis two different kinds of attractant were used in malaise traps, and insect diversity was assessed. Attractants were found to alter the dispersion of insects caught in traps. These findings can inform future malaise traps studies on insect diversity.

Here, seeking to identify a possible explanation for the more frequent diagnosis of autism spectrum disorder (ASD) in males than females, they sought to investigate a potential sex bias in the expression of ASD-associated genes. Based on their analysis, they identified 17 ASD-associated candidate genes that showed stronger collective sex-dependent expression.

Inherent bias is often the unconscious driver of human behavior, and the first step towards overcoming these biases is our awareness of them. In this article the authors investigate whether race, gender or species affect the choice of charity by middle class Spaniards. Their conclusions serve as a starting point for further studies that could help charities refine their campaigns in light of these biases effectively transcending them or taking advantage of them to improve their fundraising attempts.

The authors explore how diversity in data sets contributes to bias in artificial intelligence.

The authors investigate the effects of popular media consumption on the public's opinion on nuclear power. They find that regardless of education level or positive/negative bias of the article, participants are willing to modify their opinions on nuclear power after consuming a single article.