<\/p>\n
![\"\"](\"https:\/\/www.newsbeep.com\/us-tx\/wp-content\/uploads\/2026\/09\/6ab3181455047.image.jpg\")
<\/p>\n
Photo courtesy of Shuchisnigdha Deb<\/p>\n
Generative artificial intelligence has been used by many students since its rise in popularity with systems like ChatGPT and Google Gemini. <\/p>\n
When Shuchisnigdha Deb, associate professor in the Department of Industrial, Manufacturing and Systems Engineering, noticed this, she decided to make an AI tool that would teach students rather than just give them the answers. <\/p>\n
However, the road to this AI guide did not start at education. <\/p>\n
Deb developed and tested a version of the system with the Fort Worth Police Department before she turned to focus on students. The AI system aimed to train police officers on de-escalation tactics by mimicking the unpredictability of real-life scenarios. <\/p>\n
The use of AI removed the difficulty of scheduling conflicts with skilled roleplayers and provided the sense of uncertainty needed for the training. <\/p>\n
\u201cThere\u2019s no way for those police officers to know exactly what\u2019s wrong in that situation, and they are going into somewhere where they are also scared of their life because sometimes those people may have weapon,\u201d Deb said.\u00a0<\/p>\n
The de-escalation model took many iterations, she said. <\/p>\n
Deb and her team had to rely on open source models, AI that has underlying code and training structures for anyone to modify, as they didn\u2019t have much real scenario data to work on. <\/p>\n
They prompted the model with principles important for de-escalation training. While using the system, officers still have to build a rapport and engage in empathetic conversation to calm the situation. <\/p>\n
Deb incorporated hooks, conversational topics people in crisis may be excited to talk about, for officers to pick up on while talking. <\/p>\n
\u201cIf officers can pick up on the hook, then that means the training is going toward a positive direction, toward de-escalation,” she said. \u201cIf they cannot, we\u2019ll be giving the hook two or three times throughout the conversation.\u201d <\/p>\n
While practicing with the model, it was difficult to mimic a smooth conversation between the police officer and the model due to the time it took to convert the officer\u2019s speech to text, then turn the response into audio. It took around nine seconds for the AI to give a response back, she said. <\/p>\n
Deb and her team wanted the system\u2019s response to be as quick as a human response. She said they were excited when they were able to reduce the response time from nine seconds to two seconds or less. <\/p>\n
\u201cPolice officers are coming to our lab and going through the training, and their feedback and how excited they are about this \u2014 that\u2019s what is making us really proud of ourselves, our work,\u201d Deb said. <\/p>\n
Fort Worth police officer Jorge Lopez acted as the subject matter expert for the project, providing Deb and her team with information on how officers would typically respond in different scenarios. <\/p>\n
Lopez said he was intimidated at the beginning because he was not familiar with the technical aspects of the project. Then, as he got to know the developers, he became more comfortable. <\/p>\n
\u201cI\u2019ve come to consider them friends because they\u2019re a hardworking group of developers, and they\u2019re amazing, what they\u2019ve been able to create from scratch,\u201d he said. <\/p>\n
Lopez has been trying to compress his over 30 years of de-escalation experience into training to give recruits opportunities to get practice earlier. <\/p>\n
He said he could see the model being used for other purposes, such as tactical communications and hostage negotiation training. <\/p>\n
\u201cReally the sky\u2019s the limit on getting officers training, anytime they have downtime,\u201d Lopez said. \u201cOn your lunch break you can run a couple of scenarios, put on the headset and learn something, practice something, which is always going to be good.\u201d <\/p>\n
The collaboration with the Fort Worth Police Department was a good proof of concept for Deb\u2019s next project, she said. <\/p>\n
Now, with a $750,000 grant from the National Science Foundation, Deb aims to adapt the model tested with the Fort Worth Police Department to guide students through their assignments. <\/p>\n
\u201cWe are worried that students are losing their creativity and also the habit of learning,\u201d she said. \u201cWe are all screaming about AI taking our job, but we aren\u2019t creating a competitive workforce.\u201d <\/p>\n
The AI model will be smaller for the classroom, and it will be programmed with only the information for specific courses, she said. <\/p>\n
Deb and her team will observe students who have not used the AI model and compare their learning with that of students who have interacted with the model. <\/p>\n
\u201cThey will be more open to think critically, not just take AI’s response for granted,\u201d Deb said. <\/p>\n
She wants students to be able to use their voices and communicate properly in professional settings like job interviews and meetings instead of relying on text, she said. <\/p>\n
\u201cI\u2019m not saying that using AI is bad,\u201d Deb said. \u201cIt\u2019s a tool. If it makes you efficient, why not use it? But I want them to have the skill to understand when it\u2019s wrong, when it\u2019s correct, and even if it\u2019s writing something for us, I need them to have the ability to make the edits.\u201d <\/p>\n
@kh.kennedyharvey<\/p>\n