Michelle Vaccaro

Alexander K. Saeri, Jess Graham, Michael Noetel et al.

Artificial intelligence poses many risks, ranging from familiar present-day harms to unprecedented and potentially catastrophic ones. Effective risk management requires prioritization: we must understand which risks are most severe, who is most vulnerable, and who is most responsible for addressing them. We report results from a three-round Delphi study conducted late 2025 with 272 international AI experts. Experts rated 24 AI risks on harm probability and severity, sector and actor vulnerability, actor responsibility, and overall concern. Experts estimated the five most severe harms in the next 5 years were likely to come from dangerous capabilities, competitive dynamics, weapons & cyberattacks (including CBRNE), power centralization, and false information. In a business-as-usual scenario, experts judged 18 of 24 risks as having a more than 10% probability of catastrophic outcomes (e.g., more than 1 million deaths or more than USD 100B in financial loss) in the next 5 years (2025-2030). In a scenario where pragmatic mitigations are implemented, experts still judged five risks as having a more than 10% probability of catastrophic outcomes: dangerous capabilities, weapons & cyberattacks, environmental harm, inequality & unemployment, and power centralization. All 24 risks were judged as being more than 5% likely to cause catastrophic outcomes. AI users and the general public were judged the most vulnerable to these risks, but experts assigned the highest responsibility for addressing them to general-purpose AI developers and governance actors (including governments, regulators, and standards bodies). Across most risks, experts identified information, finance, and national security as the most vulnerable sectors. These findings can guide AI risk prioritization and clarify expert expectations about who should bear responsibility for mitigation.

Andres Campero, Michelle Vaccaro, Jaeyoon Song et al.

The Turing test for comparing computer performance to that of humans is well known, but, surprisingly, there is no widely used test for comparing how much better human-computer systems perform relative to humans alone, computers alone, or other baselines. Here, we show how to perform such a test using the ratio of means as a measure of effect size. Then we demonstrate the use of this test in three ways. First, in an analysis of 79 recently published experimental results, we find that, surprisingly, over half of the studies find a decrease in performance, the mean and median ratios of performance improvement are both approximately 1 (corresponding to no improvement at all), and the maximum ratio is 1.36 (a 36% improvement). Second, we experimentally investigate whether a higher performance improvement ratio is obtained when 100 human programmers generate software using GPT-3, a massive, state-of-the-art AI system. In this case, we find a speed improvement ratio of 1.27 (a 27% improvement). Finally, we find that 50 human non-programmers using GPT-3 can perform the task about as well as--and less expensively than--the human programmers. In this case, neither the non-programmers nor the computer would have been able to perform the task alone, so this is an example of a very strong form of human-computer synergy.

Michelle Vaccaro, Jaeyoon Song, Abdullah Almaatouq et al.

Current frontier AI safety evaluations emphasize static benchmarks, third-party annotations, and red-teaming. In this position paper, we argue that AI safety research should focus on human-centered evaluations that measure harmful capability uplift: the marginal increase in a user's ability to cause harm with a frontier model beyond what conventional tools already enable. We frame harmful capability uplift as a core AI safety metric, ground it in prior social science research, and provide concrete methodological guidance for systematic measurement. We conclude with actionable steps for developers, researchers, funders, and regulators to make harmful capability uplift evaluation a standard practice.

Michelle Vaccaro, Michael Caosun, Harang Ju et al.

We conducted an International AI Negotiation Competition in which participants designed and refined prompts for AI negotiation agents. We then facilitated over 180,000 negotiations between these agents across multiple scenarios with diverse characteristics and objectives. Our findings revealed that principles from human negotiation theory remain crucial even in AI-AI contexts. Surprisingly, warmth--a traditionally human relationship-building trait--was consistently associated with superior outcomes across all key performance metrics. Dominant agents, meanwhile, were especially effective at claiming value. Our analysis also revealed unique dynamics in AI-AI negotiations not fully explained by existing theory, including AI-specific technical strategies like chain-of-thought reasoning, prompt injection, and strategic concealment. When we applied natural language processing (NLP) methods to the full transcripts of all negotiations we found positivity, gratitude and question-asking (associated with warmth) were strongly associated with reaching deals as well as objective and subjective value, whereas conversation lengths (associated with dominance) were strongly associated with impasses. The results suggest the need to establish a new theory of AI negotiation, which integrates classic negotiation theory with AI-specific negotiation theories to better understand autonomous negotiations and optimize agent performance.

Michelle Vaccaro, Abdullah Almaatouq, Thomas Malone

Inspired by the increasing use of AI to augment humans, researchers have studied human-AI systems involving different tasks, systems, and populations. Despite such a large body of work, we lack a broad conceptual understanding of when combinations of humans and AI are better than either alone. Here, we addressed this question by conducting a meta-analysis of over 100 recent experimental studies reporting over 300 effect sizes. First, we found that, on average, human-AI combinations performed significantly worse than the best of humans or AI alone. Second, we found performance losses in tasks that involved making decisions and significantly greater gains in tasks that involved creating content. Finally, when humans outperformed AI alone, we found performance gains in the combination, but when the AI outperformed humans alone we found losses. These findings highlight the heterogeneity of the effects of human-AI collaboration and point to promising avenues for improving human-AI systems.