DETECT Scam Detection & Prevention
Detecting behavioral patterns of manipulation, scams, and emerging AI-enabled harms.
Overview
DETECT develops behavioral approaches to identify psychological manipulation, scams, and emerging AI-enabled harms before significant harm occurs.
Rather than relying primarily on suspicious keywords or isolated messages, we examine how manipulation develops across conversations—through patterns of trust-building, persuasion, emotional influence, financial probing, and escalation.
This work supports the development of behavioral detection methods that can help platforms, financial institutions, governments, and other organizations identify emerging risks earlier and intervene more effectively.
Global Initiative
This work also forms the foundation of the Global Partnership on AI-Enabled Manipulation & Scams, an international initiative selected by the United Nations AI Dialogue Partnerships Hub to advance research, evaluation methods, and cross-sector collaboration on AI-enabled manipulation and fraud.
Research Approach
Our research is grounded in a proprietary multilingual dataset of more than 175 documented scam cases and over 15,000 scammer messages collected from real-world interactions.
We analyze these conversations to identify behavioral patterns, persuasion strategies, cognitive mechanisms, and escalation pathways used in sophisticated scams.
These insights inform the development of a Behavioral Manipulation Taxonomy, specialized detection models, multilingual datasets, and tools designed to identify risk earlier in an interaction.
Behavioral AI Lab develops the underlying behavioral frameworks, methodologies, and research, and works with technical collaborators to translate these insights into scalable detection systems.
Expected Inpact
DETECT aims to shift scam prevention from reacting after financial loss to identifying behavioral risk earlier in the interaction.
The work is designed to support financial institutions, technology platforms, governments, and other organizations seeking to protect people from increasingly sophisticated forms of digital manipulation and fraud.
By combining behavioral science with AI-based detection, we aim to create scalable systems that can recognize emerging manipulation patterns across languages, platforms, and evolving forms of AI-enabled scams.
DETECT Scam Detection & Prevention
Detecting behavioral patterns of manipulation, scams, and emerging AI-enabled harms.
Overview
DETECT develops behavioral approaches to identify psychological manipulation, scams, and emerging AI-enabled harms before significant harm occurs.
Rather than relying primarily on suspicious keywords or isolated messages, we examine how manipulation develops across conversations—through patterns of trust-building, persuasion, emotional influence, financial probing, and escalation.
This work supports the development of behavioral detection methods that can help platforms, financial institutions, governments, and other organizations identify emerging risks earlier and intervene more effectively.
Global Initiative
This work also forms the foundation of the Global Partnership on AI-Enabled Manipulation & Scams, an international initiative selected by the United Nations AI Dialogue Partnerships Hub to advance research, evaluation methods, and cross-sector collaboration on AI-enabled manipulation and fraud.
Research Approach
Our research is grounded in a proprietary multilingual dataset of more than 175 documented scam cases and over 15,000 scammer messages collected from real-world interactions.
We analyze these conversations to identify behavioral patterns, persuasion strategies, cognitive mechanisms, and escalation pathways used in sophisticated scams.
These insights inform the development of a Behavioral Manipulation Taxonomy, specialized detection models, multilingual datasets, and tools designed to identify risk earlier in an interaction.
Behavioral AI Lab develops the underlying behavioral frameworks, methodologies, and research, and works with technical collaborators to translate these insights into scalable detection systems.
Expected Inpact
DETECT aims to shift scam prevention from reacting after financial loss to identifying behavioral risk earlier in the interaction.
The work is designed to support financial institutions, technology platforms, governments, and other organizations seeking to protect people from increasingly sophisticated forms of digital manipulation and fraud.
By combining behavioral science with AI-based detection, we aim to create scalable systems that can recognize emerging manipulation patterns across languages, platforms, and evolving forms of AI-enabled scams.