DHS embraces GenAI’s quirks to train asylum and refugee officers, simulating realistic interviews and unpredictable scenarios.

The Department of Homeland Security (DHS) has launched a GenAI pilot program to train new asylum and refugee officers by simulating realistic interview scenarios. This innovative approach leverages GenAI to create unpredictable and dynamic responses, closely mirroring actual conversations with asylum seekers.

Michael Boyce, director of the DHS AI Corps, explained that the GenAI system generates answers to interview questions posed by officers-in-training. The AI simulates refugee applicants, providing varied and sometimes inconsistent responses that reflect real-life interviews. This method allows trainees to practice conducting three-hour interviews without the need for experienced personnel to play the role of refugees.

Boyce highlighted that embracing GenAI’s tendency to produce less predictable and occasionally inaccurate responses is intentional. This approach helps replicate the complexities of real interviews, where officers often deal with interpreters and applicants who may provide confusing or incomplete information. By training with GenAI, officers learn to navigate these challenges effectively.

The pilot aims to enhance training by generating dynamic, personalized materials that adapt to the specific needs of officers. By exposing trainees to a wide range of scenarios, including sensitive and traumatic topics, the GenAI system ensures comprehensive preparation for handling real-world cases. DHS Secretary Alejandro Mayorkas noted that the AI is also being taught to be reticent, mirroring the hesitancy of refugees to discuss their traumatic experiences.