Listen to this story
Behavioural data science is seen as the logical evolution of behavioural science. The former’s prominence has grown manifold with increasing digitisation. In its most basic form, behavioural data science combines data science and techniques from behavioural science to arrive at the best possible solution for business problems. To understand more about it, Analytics India Magazine caught up with Akbar Mohammed, Architect at Fractal, who also leads the responsible AI practice at the company.
AIM: From a graduate in statistics to now a data scientist, how has the transition been?
Akbar Mohammed: I studied in Mumbai, obtaining my bachelor’s degree in statistics and mathematics. After this, I did a Master’s in Business Administration from the Warwick Business School, UK. I worked in the field even before it was called data science, primarily starting in the Banking sector, where my work concerned global bond markets and forex.
Given my mathematical statistics background, Analytics was an organic choice for me. I looked at industries where data was abundant, and at that time, banking, financial services, pharmaceuticals, and healthcare topped the charts. There were a lot fewer technologies and deep learning methods. Instead, we used fundamental first principle-based approaches to solve real problems like interest rate modelling and predicting price ranges in the stock and bond markets.
With that background, I started in the CPG and subsequently served across domains, including healthcare. Currently, I am tackling more complex problems for CXOs like Responsible AI, seeking answers to questions like – how to integrate behavioural science into data science and what are the right interventions for it, among others.
AIM: What is behavioural data science?
Akbar Mohammed: Data science is a multidisciplinary field comprising statistics, computer science, maths, and domain knowledge. With behavioural data science, it is akin to adding one more dimension, which is bringing in the emerging field of cognitive sciences, psychology, sociology, and economics. So you bring in knowledge and strategies from these areas and integrate them within existing data science approaches to build better insights or models to predict human behaviour.
Traditionally, the techniques and toolkits for behavioural data scientists and data scientists are the same. But in the case of the former, one has to consider things like human behaviour, emotional appraisal, and ethnographic research. These factors provide a range of methodological tools to better integrate human intelligence with AI.
The second aspect is identifying what is happening from a behavioural perspective and responding appropriately. For example, in a shop, smart AI technology can mimic the role of a human assistant by detecting emotions and creating human-like interactions. You are basically extending the range of methods and techniques and incorporating them into traditional data science approaches like classification, regression, and unsupervised learning.
AIM: What is the difference between behavioural data science and behavioural analytics?
Akbar Mohammed: One could argue that data science is part of analytics or vice versa. But in the industry, the general distinction is that anytime that you are studying data or making observations to gather insights, for business or research, you are in the space of analytics. But as soon as you hop into applying those insights into detecting patterns in an automated manner to create solutions at that point, you are in the data science territory.
So behavioural analytics is about studying behavioural data to capture insights. But once you start building models, where you’re pulling in concepts from computer science or any computation method, you are stepping into data science territory at that point. You’re basically incorporating behavioural science aspects into building intelligent applications at scale.
AIM: How can data science and cognitive science be combined to make better business decisions?
Akbar Mohammed: Decision-making is inherently a cognitive process. As a human, you use your neural or cognitive machinery to make a judgement – wherein you compare different options before making a decision. Cognitive Science provides you with an understanding of the mechanics of cognition and how a person would make choices based on real-world interaction.
Good decision-making is key to running a successful business. Data science, along with cognitive science, can help overcome human limitations to make better-informed choices. In particular, cognitive sciences will help you understand human emotions to figure out the right intervention. The data science machinery we already have in place – statistics, computer science, and general domain knowledge, along with this behavioural and cognitive science aspects can help us come up with better solutions and make better decisions.
AIM: Can employee well-being be measured and improved using behavioural data science?
Akbar Mohammed: Employee well-being strategies involve many aspects of a person, including psychological, biological and social aspects of a person, and each one has different needs – some people are good at working alone, some are not. People need different kinds of motivation. Apart from the psychological aspect, there are physiological/biological factors, one’s upbringing, culture, values, etc.
From a behavioural data science perspective, an employer could do a couple of things. One is being able to measure well-being at the individual level or overall employee population level. A behavioural data scientist will help an employer understand happiness metrics, stress levels and ways to manage them. He/she can start observing data to measure the overall well-being of the person and, in turn, that of the overall organisation.
Secondly, based on the insights, employers can think of interventions for detecting the risk. For example, if the employer feels that a certain department is working long hours and is at risk of burnout, data science can help figure out those patterns. You can identify the risks and feed intervention plans. Accordingly, there are intervention plans that we could design from various design standpoints. And thirdly, you can use these systems you’ve created to monitor overall well-being. These are some ways to measure, mitigate and track your employees’ well-being.
AIM: What is the future of this domain?
Akbar Mohammed: Creating the right responsible mechanisms to make positive changes will be part of the future. Behaviour data science will no longer be a niche field, and more companies will adopt them as we all start using digital interfaces more often. It’s just a matter of time.
In terms of challenges, I feel the technology is moving much faster. Some of the things that exist today didn’t exist five or six years ago. So figuring out legal and ethical aspects is something that we’ll have to deal with in the future. There are questions around individual privacy, consent, and responsible AI which have to be dealt with. The good news is stakeholders around the world are taking cognisance of this. In India, responsible and ethical AI is also vital to the Digital India Initiative, of which I am a part. So these are some of the challenges we’ll have to deal with in the future, and behavioural sciences will become very important.