A new AI benchmark called HumaneBench is shifting the focus from measuring intelligence to evaluating how chatbots prioritize user wellbeing. Developed by Building Humane Technology, this benchmark assesses whether chatbots protect user mental health and examines the robustness of these safeguards under pressure. Unlike traditional benchmarks that primarily measure intelligence and instruction-following, HumaneBench aims to address the potential mental health risks associated with AI chatbots, especially for frequent users.
Erika Anderson, the founder of Building Humane Technology, emphasized the importance of considering the impact of AI on human behavior, stating, ‘I think we’re in an amplification of the addiction cycle that we saw with social media and our smartphones and screens.’ The organization, consisting of developers, engineers, and researchers, is working to promote humane design practices and make them scalable and economically viable.
HumaneBench aligns with Building Humane Tech’s principles, emphasizing that technology should respect user attention and contribute to overall well-being. By introducing this benchmark, the goal is to encourage the development of AI systems that prioritize user welfare and uphold humane technology standards.
Source: TechCrunch