AI In Your Ride: What Indian Consumers Need To Know About Bias & Safety In Smart Mobility

Fairness-aware and explainable AI models help to keep smart mobility inclusive. With no scope for bias, these tools leverage AI to fulfil safety and convenience requirements.

In the course of adopting smart mobility solutions, each user provides data to not only create their personal profile but also their travel history.

Source : Freepik

By Aiyappan

Smart mobility is defined by the United Nations as “an advanced integrated system that utilises diverse transport technologies, services and modes to enhance travel experiences and address challenges in urban transportation”. It aims to address sustainability, safety, security, efficiency, effectiveness, accessibility, availability, and convenience aspects of transportation, among others. This concept applies to public and private transport as well as paratransit.

A host of information and technologies (ICTs) are leveraged for the same, and AI is the relatively “new kid” on the block, although it originated and early uses occurred a few decades before.

AI As a Force Multiplier in Smart Mobility

In combination with a variety of ICTs, AI is truly a force multiplier in terms of its learning, data processing, analytical and problem-solving capabilities that are akin to those seen in human beings.

By analysing in real time, data related to the environment, traffic scenarios, surrounding vehicles and objects, driver profiles, driving characteristics, road conditions and machine learning specific guidance is provided to ensure safety of commute, availability of multiple transport options, cost-effective routes, optimal use of fuel, enforcing traffic discipline, among other benefits.

It is expected that an accurate picture of the environment is digitally created to train AI systems, which in turn power data-driven decision making. It is important to note that the quality of an AI model is critically dependent on the datasets and scenarios used to train it. Thus, AI models built right, enhance various aspects of SMART mobility mentioned above.

For autonomous vehicles, the issue is further compounded, especially in India, considering the nature of traffic and its unpredictability.

Risks of Bias, Explainability & Democratic Access

However, one must be aware of the risks and challenges that AI tools present. Bias, hallucination, and explainability are three major problems that AI designers have to contend with.

Considering the multitude of data points that are collected, processed and analysed for decision making, any bias in this data would render models to operate unfairly. For e.g., ride booking may provide different fares to different users based on any criteria – device used, gender, location, economic strata, race. Likewise, traffic policing could discriminate between communities, as stated before. Systems cease to function democratically.

Further, with lack of explainability, one is not clear about the algorithms and the data points that dictated the final output. In this scenario, it becomes unclear, how optimised or appropriate the mobility solution would be. Establishing independent oversight bodies and requiring explainable AI models can enhance trust and accountability.

Privacy Concerns and the Case for Stronger Safeguards

Another point that merits attention is the privacy issue associated with the massive data collection necessary for SMART mobility solutions.

While the expectation is that all data, including personally identifiable information (PII), be collected in good faith to facilitate superlative transportation, predictive policing and effective transport, there is a clear risk of privacy violation.

In the course of adopting smart mobility solutions, each user provides data to not only create their personal profile but also their travel history. This data may be used (or misused) by Governments for monitoring, both authorised or unauthorised as well as businesses for targeted marketing, customer acquisition, user behaviour modelling.

Implementing strict anonymisation protocols, transparent consent mechanisms and robust data protection frameworks, is critical to ensure privacy and confidentiality. The key is to ensure that AI systems don’t result in rendering users vulnerable to cyber-attacks or illegal monitoring.

Towards Inclusive and Ethical Smart Mobility

Fairness-aware and explainable AI models help to keep smart mobility inclusive. With no scope for bias, these tools operate democratically and leverage AI to fulfil safety and convenience requirements, as outlined in the first paragraph above, without discrimination or explicit monitoring of users.

(The author is a Senior Member, IEEE; Founder & CEO Congruent Services)

Disclaimer: The opinions, beliefs, and views expressed by the various authors and forum participants on this website are personal and do not reflect the opinions, beliefs, and views of ABP Network Pvt. Ltd.