New research from the MIT Sloan School of Management identifies the most effective ways businesses can use and benefit from AI personalization
Personality Pairing Improves Human-AI Collaboration
Cambridge, MA, Aug. 31, 2026 (GLOBE NEWSWIRE) — While Artificial Intelligence (AI) is receiving increased attention for its potential to improve productivity and performance, the results can vary widely across different tasks, industries, and individuals. Developers can now shape AI personalities, with the goal of customizing AI agents for individual users, but most previous research into the benefits of AI have used experiments providing all participants with identical AI tools; however, a key question remained as to whether variations in performance could be related to the tailoring of AI agents for their human collaborators.
A new paper published in the Proceedings of the National Academy of Sciences (PNAS), “Personality Pairing Improves Human-AI Collaboration” presents the first-ever, large-scale, randomized controlled experiment testing how AI agent “personalities” interact with human personalities to shape human-AI collaboration and performance, and what combinations proved to be the most effective.
Co-authors are Sinan Aral, MIT Sloan School of Management professor and director of the MIT Initiative on the Digital Economy; and Harang Ju, an assistant professor at the Johns Hopkins University Carey Business School. They are the co-founders of the technology company Pairium AI, which is developing AI personalization technology to optimize human-AI collaboration.
“To date, research on human teams has consistently shown the advantages of bringing certain, complementary personalities and abilities together,” said Aral. “Our new research shows that we can implement personalized AI in much the same way if you match the right type of personalization of the AI agent to the individual, which maximizes performance on the task.”
Which types of AI personalization work best?
Aral and Ju conducted an experiment involving more than 1,200 participants from the U.S. using Pairit, a platform they developed to randomize human-AI collaborations. The human participants were randomly paired with AI agents prompted to independently exhibit high or low levels of the Big Five personality traits: openness, conscientiousness, extraversion, agreeableness, and neuroticism.
The human-AI teams worked together to create display ads marketing a year-end report of a real think tank during a 40-minute session using real-time chat and synchronized text- and image-editing tools. Researchers measured the quantity of ads they were able to produce within this time frame, and then a separate group of approximately 1,100 people rated the ads on image and text quality. The ratings revealed that some types of AI personalization yielded better outcomes than others.
“When it comes to working with AI, we find that very specific pairings can make a significant difference in terms of performance,” says Ju.
The most highly rated text and/or images were created when extraverted humans worked with extraverted AI, conscientious humans worked with conscientious AI, and open humans and conscientious AI. Conversely, certain pairings of human and AI personalities appeared to significantly reduce ad quality, including extraverted humans and conscientious AI, conscientious humans and agreeable AI, and neurotic humans and conscientious AI.
What are the implications of personalizing AI for businesses?
To better understand the actual benefits of ad quality and performance, the researchers then ran the ads on the social media platform X for a two-week period, during which the ads received approximately five million impressions. They measured both click-through rate and cost per click, finding that the quality of the ads, as reflected in their ratings, predicted their real-world performance. They found that text quality increased the click through rate by about 7% and decreased the cost per click by about 30 cents per click. This highlights the opportunity for training AI to create better text and images to obtain substantial marketing results.
“In the first part of our research, people had reported that the better performing ads were higher quality overall—with higher image and text quality—and were generally just better made,” says Aral. “Then we were able to see that play out in the real world when they performed better in the advertising market.”
Aral and Ju are planning further research in other settings and industries including sales, engineering, and healthcare.
“We believe everyone can benefit from a high level of personalization for their own agent use by understanding the types of personality pairings that work best across settings and industries,” said Ju.
Attachment
- Personality Pairing Improves Human-AI Collaboration
CONTACT: Casey Bayer MIT Sloan School of Management 914.584.9095 [email protected] Patricia Favreau MIT Sloan School of Management 617.895.6025 [email protected]
