
CHENNAI: Artificial intelligence (AI) continued to dominate discussions on Day 2 of the ThinkEdu Conclave 2025, presented by SASTRA University. Prof Venkatesh Balasubramanian, Head of RGB (Rehabilitation Bioengineering Group) Labs at IIT-Madras, captivated the audience with insights into how AI-driven data and predictive models have the potential to shape human behaviour.
Speaking in the session “AI and Behavioural Change: A Roadmap”, he noted that specific human behaviours such as checking horoscopes, going on shopping sprees, being wary of self-driving cars, and trying to misuse government schemes would all be impacted by AI.
He described various challenges in building workable AI models, the most prominent of them being the lack of availability of reliable data. He added that data is often susceptible to the biases and affiliations of the sources, which could result in inaccurate results.
Other challenges he mentioned include the embargoes on hardware to analyse AI models, without which they end up becoming “beautiful shooting in the dark”, unfair and flawed hypotheses for AI models, and the ethical and legal conundrums AI finds itself tied in.
The professor also mentioned a phenomenon among several AI researchers, including himself — that of trading a model’s accuracy with its ability to accomplish things.
Moulding human behaviour through data & predictive models
Talking about his work with the Union Ministry of Road Transport and Highways (MoRTH) through his think tank the Centre of Excellence for Road Safety (CoRS), Prof Balasubramanian says that patterns of human behaviour can be tracked and changed by analysing data of their consequences.
He responded to questions by senior journalist Kaveree Bamzai on the initiative and said that the Centre’s data analytics model helped reduce the number of road accidents in Tamil Nadu last year.
Explaining the process, he said that the model’s analysis of the road fatality figures from Tamil Nadu showed that the nights of December 31 showed a high number of incidents every year, and these accidents occurred on junctions where high-speed roads met low-speed roads. The dominant demographic involved in these accidents was young males.
“These men could not afford to attend parties, so they rent bikes and take out rallies on the roads. The first step was to offer them alternatives to these activities, which they could enjoy,” he explained.
He added that these insights also enabled the municipal administrations to increase vigilance at junctions to monitor traffic better.
The professor cautioned that human behaviour may remain unchanged despite solid data that could influence change. “For example, head injuries being fatal is a hard fact. This still does not make some riders wear their helmets,” he stated.
While AI models can map human behaviour and predict it, changing it is where the real challenge lies, said Professor Balasubramanian.
“Data is post-facto. We need to build models to ensure that events like this don’t happen,” he asserted.