‘Accident is not act of god, but act of humans, AI can only help in getting better data and analysis

How does the Sanjaya-Road Safety Super App work?

As the name implies, Sanjaya is designed to give battlefield intelligence to the district bureaucrats. Data is collected, analysed, and then presented in a way that users can easily utilise. Just like iRAD, a single umbrella for accident data, Sanjaya is a single umbrella for utilisation. In this, a variety of AI models are used for the analysis and that analysis is simplified so that you can visualise what is the data that are there, have reports and also track any interventions that are planned.

What are the key parameters that it analyses for road safety?

We get very good location data in iRAD (Integrated Road Accident Database). So, Sanjaya gives a good analysis of the location information. It gives information on black spots or accident-prone areas in the district. It also collects perception data. We have something called a field perception survey, where local knowledge is captured. People on the ground often know there’s a possibility of an accident at a certain spot even if nothing major has happened yet. That kind of insight is collected. Local police officers or beat constables know these risky locations, but unless there’s a serious accident, it usually doesn’t get formally recorded. So we have another application through which field officers can share this perception data. We then analyse that information and utilise it. The app also indicates the preparedness level of trauma centres. The AI system assesses how prepared a trauma centre is to handle accident cases. The health department gives information in this regard such as location of the trauma centre, equipment availability, human resource etc.

Has there been any tangible impact of the app so far?

I wouldn’t necessarily call it tangible. What it should really do is change the mindset — that decisions should be taken based on data. That itself is the tangible shift: district leadership is starting to think differently now. We are implementing it through a programme called DDHI-Data Driven Hyperlocal Intervention. State governments like Madhya Pradesh, Telangana, Kerala, Gujarat, Karnataka and Odisha have already come forward to implement it at another level, where they want to take things further and get more done.

What about Uttar Pradesh, which sees maximum fatalities?

In Uttar Pradesh, we have not moved forward much. We will also have to try. It is not that everyone else is not going to do it. It is just that we have not been successful. It is not their fault. It is our fault.

Does it require high expenditure to implement the app?

No. It gives a framework and free analysis. With this, you don’t need a consultant. And the government of India is also creating another scheme Sadak Suraksha Mitra, which will be to provide assistance to the district administration to execute various things. So, it gives a formal structure for road safety.

Since iRAD data is still not foolproof, is there any step being taken in this respect?

The first phase of iRAD was to bring every police station, in every district and every state, under one umbrella. It involves four state subjects — health, police, transport, and the road-owning agencies. That part was completed in about 18 months. Now the second phase is focused on improving the quality of the data. For that, various assessments are being carried out, things like promptness of data entry, completeness of data, adequacy of the information, and making sure there are no errors or inconsistencies in the data. iRAD, by definition, is meant to capture all accidents. In Crime and Criminal Tracking Network & Systems (CCTNS), you mostly get accidents where an FIR has been registered. But there are many accidents where no FIR gets filed, and those should also be captured in iRAD. Ideally, the total number of accidents recorded in iRAD should be higher than what’s reported in CCTNS. The day that happens, it will be more reliable data.

Despite all these efforts, why are the fatalities still increasing?

At the beginning of iRAD implementation, states used to ask that will the accidents come down with this. My only answer was no, it will go up. Because what was not recorded earlier will now get recorded. The bigger challenge is not getting into an accident, but dying after an accident. Fatalities will go down when people become responsible. Fatalities will not go down because the government started a program. What we’ve essentially done with iRAD is like the role of Chitragupta in mythology — quietly collecting all the information. And then Dharmaraja, which in this case would be the government or district authorities, interprets that information and makes decisions.

So what is hampering the road safety efforts?

The biggest problem is that many well-meaning people genuinely want to help. But once they enter the system, it sometimes becomes more like a business. Then decision-makers get confused about which programme to follow. That multiplicity of programmes ends up complicating decision-making. Second, a proper systems approach is often missing. What we try to do is bring everyone into the process — whether it is decision-making, interventions, or data collection. That coordinated systems approach is very important.

Third, behavioural change has to happen. Accidents don’t happen because of an act of god, they happen because of human actions. So the human factor is critical, but changing behaviour takes time.

We did research on how people judge speed at night — whether an approaching vehicle is a bike, two bikes together, a car, or a truck, and how fast it is coming. Even students from IIT make mistakes. Evolutionarily, we have never been used to judging such high speeds, so our brains are still learning. That adaptation will take time. In the West, this transition happened gradually — from horse-driven transport to steam vehicles and eventually to high-speed cars like Bugatti. But in India, we’ve essentially moved from carts to very fast vehicles within just a few decades. So it’s not just about administration, it is also about human adaptation and evolution.