Using AI to address inequality and inclusivity in economics: Introducing Tamoghna Halder

Friend of the Stone Centre Tamoghna Halder is an accomplished columnist and an Assistant Professor of Economics at Azim Premji University in Bengaluru, India.

‍

Tamo is developing an AI tool to help first-generation students from disadvantaged backgrounds studying economics, using CORE’s The Economy as a base text. Tamo recently stopped by the Stone Centre at UCL where we caught up to find out more about the inspiration behind his AI tool.

‍

Hi Tamo, thank you for joining us. What led you to start thinking about using AI for teaching economics?

‍

My research interests as an economist lie on the intersection of economic history and caste-based inequality in India. In India, the caste system has existed for centuries and continues to have an impact today. One major way in which caste continues to contribute to learning gaps is through access to English as a language. I wondered if AI could be leveraged to close such gaps, which may in turn reduce larger inequalities in educational opportunities and educational access that are perpetuated by the caste system. That is my broad motivation for anything that I pursue on the front of AI, including the current tool.

‍

What led you to start developing your AI tool?

‍

Over the last two years, I observed a small number of academics who thought AI would solve the world’s problems, and then a large chunk who remained in denial, while others focussed on how to combat AI-enabled cheating. I find each approach to be problematic. Instead, I asked what I could practically do with AI that’s not necessarily adversarial, and how to rebuild the trust between a teacher and the student in this environment of doubt.

‍

I don’t have a strong coding background. It was a daunting task to jump into this world. So instead of thinking of a tool to serve thousands of people across the globe, I first considered what AI could do for my classroom.

‍

I recognised that students could turn to AI for help with even the simplest data analysis tasks, but not all students would truly learn from this if they just copy-paste AI-written codes. I saw an opportunity to develop a tool that would integrate AI with R (a statistical software), so that before the student goes to AI, AI comes to them – and sends a signal to the student that the instructor wants them to get help from AI in a guided manner. Say you’re stuck running a code somewhere, suddenly a window pops up to let you know you’ve made a mistake, what kind of mistake it is and how you might code differently to avoid this. I felt this limited environment, where the teacher retains some control over the interaction between the student and AI, would be a sweet spot. This was my starting point.

‍

Do students from marginalised backgrounds have difficulty asking for help when they make mistakes?

‍

Absolutely. At least 50% of the student body at Azim Premji University are not from urban India. There are many embedded structural problems that lead to these students being behind in English language comprehension, which can make things like understanding a basic journal article a daunting task.

‍

This 50%, who did not have access to certain cultural contexts and practices, suddenly find themselves in a classroom where 50% are from top, elite urban English-taught schools who are way ahead in terms of reading and comprehension. Such situations may turn quite intimidating for many students.

‍

For example, The Economy: A South Asian Perspective Unit 4.2 discusses growing rice and cassava. Now, what is cassava? This is not a term you will grow up hearing in India. Note that this is not a conceptual struggle for the student; it’s just ‘what exactly does this word mean?’. From a caste perspective, it may be the case that an elite, dominant caste student from an urban background has heard of cassava because they have come across the word in some novel or newspaper article. In fact, even if they don’t know it, they are more likely to ask the instructor, “what is cassava?”, while the student from a marginalised caste may not ask the same out of fear of being judged or an internalised lack of confidence.

‍

Did you experience these difficulties yourself?

‍

Not inside the classroom as such, as I came from a relatively privileged section of society and had the confidence to ask questions, but when I first picked up an economics textbook, I did not understand what this thing called ‘hamburgers’ was that the book kept talking about. There was never a food item that I could relate to, and I am sure many of my friends were in the same boat. My immediate questions were often not conceptual but simply vocabulary and so-called “general knowledge”. I was therefore either searching on the internet or opening a dictionary to study economics!

‍

How does your AI tool help with this?

‍

The tool I’m building primarily helps a student in clarifying conceptual problems in the context of the textbook’s content. This will allow instructors to moderate a student’s interaction with AI for conceptual understanding. But more importantly, it will also help with clarifying language and factual comprehension that may not be immediately culturally available to you. The conversation with AI can be logged to generate reports for instructors who can then tailor their delivery in the classroom as well as in supporting students outside of the classroom through office hours and language support. The tool also brings students back to the habit of reading, as it forces one to read small chunks of text carefully and answer quizzes based on these paragraphs before they can move ahead.

‍

That's the amazing thing about AI; you can ask it anything without the fear of being judged. Even though the instructor is well-meaning, friendly, and assures you they will not judge you, caste-based differences in cultural exposures create social barriers that are often extremely difficult to overcome. AI does not care whether you know the meaning of something, so we're much more comfortable with that interface. I'm trying to leverage it to reduce compounded inequalities for marginalised-caste students and improve their long-term learning outcomes, or at least ensure a level playing field.

‍

Why did you choose CORE’s The Economy as the text basis?

‍

First, CORE’s The Economy is connected to a wide network of students and universities around the world, in many different contexts. As a base text, it offers really a nice variation in different ways for the tool to be used across these contexts. This helps me get better feedback on how useful the tool is in varying scenarios. Second, as an Economist I find that CORE’s The Economy is less ideologically biased than most introductory textbooks, so I cannot really think of a better base text.

‍

Thank you for your time, Tamo. It was great to meet you, and we look forward to hearing more about the development of your AI tool in the future!

‍