PREVENT Behavioral Interventions
Designing behavioral interventions and safer AI systems to reduce harm before it occurs.
Overview
PREVENT translates behavioral science into practical interventions and design strategies that help organizations reduce AI-related risks before harm occurs.
We examine how product design, user experience, warnings, friction, choice architecture, and other behavioral interventions can influence how people interact with AI and digital systems.
The goal is to help organizations build environments that support safer decisions, appropriate trust, and more resilient human–AI interactions.
Research Approach
Our approach combines behavioral science, experimentation, and Safety by Design to identify where interventions can reduce risk across the user journey.
We examine how people respond to AI-generated content, recommendations, warnings, interfaces, and conversational systems, and use behavioral insights to design interventions that can be tested in real-world environments.
This may include changes to choice architecture, warning design, friction, information presentation, user controls, and other product features intended to improve decision-making and reduce harmful outcomes.
Expected Inpact
PREVENT aims to move AI safety from identifying risks to actively reducing them.
By embedding behavioral science into product and system design, organizations can intervene earlier, help users recognize risk, reduce susceptibility to manipulation and overreliance, and support safer decision-making.
The long-term goal is to make behavioral safety a practical part of how AI products are designed, tested, and deployed.
PREVENT Behavioral Interventions
Designing behavioral interventions and safer AI systems to reduce harm before it occurs.
Overview
PREVENT translates behavioral science into practical interventions and design strategies that help organizations reduce AI-related risks before harm occurs.
We examine how product design, user experience, warnings, friction, choice architecture, and other behavioral interventions can influence how people interact with AI and digital systems.
The goal is to help organizations build environments that support safer decisions, appropriate trust, and more resilient human–AI interactions.
Research Approach
Our approach combines behavioral science, experimentation, and Safety by Design to identify where interventions can reduce risk across the user journey.
We examine how people respond to AI-generated content, recommendations, warnings, interfaces, and conversational systems, and use behavioral insights to design interventions that can be tested in real-world environments.
This may include changes to choice architecture, warning design, friction, information presentation, user controls, and other product features intended to improve decision-making and reduce harmful outcomes.
Expected Inpact
PREVENT aims to move AI safety from identifying risks to actively reducing them.
By embedding behavioral science into product and system design, organizations can intervene earlier, help users recognize risk, reduce susceptibility to manipulation and overreliance, and support safer decision-making.
The long-term goal is to make behavioral safety a practical part of how AI products are designed, tested, and deployed.