Crucial roles for social sciences in AI, says Sergei Guriev
He told University World News in an interview: “The political and social implications of AI are directly related to the ethics of AI. A lot of social scientists educate computer science colleagues to think about what they invent and how they invent it.
“Economists and sociologists who work on inequality, for example, can teach AI colleagues how to develop AI algorithms that do not reproduce discrimination and bias. This is a huge issue.”
He gave the example of using an algorithm to sort through the CVs of applicants for a job. That is doable by looking at which candidates with which CVs were successful in the past, and then training an algorithm using that data set to sort through CVs in future. The problem is that the algorithm will reproduce the flawed human choices of the past.
“The idea that a machine can replace a biased, discriminating human, and fix the discrimination problem, is a utopia – unless we involve social sciences, understand the biases and correct them at the level of the algorithm,” he said.
“Universities must not only teach how to develop algorithms in terms of maths or computer science, but also to think about the social implications of AI at the development stage, so that we do not reproduce bad things from the past and we do not face a backlash against technology.”
A ChatGPT hullabaloo
Controversy engulfed Sciences Po – otherwise known as the Paris Institute of Political Studies – following a 27 January email that Guriev wrote to students, saying they were not allowed to use ChatGPT without explicit reference to it.
“While it was interpreted by French media as Sciences Po prohibits ChatGPT, the real news was that we banned students from using ChatGPT without acknowledging in their work that they had done so. This is a very straightforward move,” Guriev explained. Sciences Po views plagiarism as pretending to write something that is not yours – even if it is stolen from a robot, not a person.
“But otherwise we encourage the use of ChatGPT, including to understand how it works. We also encourage the use of ChatGPT in teaching and learning and with the supervision of faculty. But we are a free country, so if you want to use ChatGPT and then tell us, ‘I wrote this paragraph with the help of ChatGPT’, that’s fine, that’s absolutely normal.”
Guriev believes that with all the hype around ChatGPT, the big problem that it does not always produce correct answers and can be misleading, got somewhat lost. “So it is a tool that we need to learn how to use. This tool will change itself, but overall it is just a tool to support humans. It’s not a replacement for a human.”
The global hullabaloo around ChatGPT, released in November 2022, prompted the baptism of fire for Guriev, who was appointed provost of Sciences Po in July 2022. But he has faced worse.
Guriev obtained an MSc summa cum laude in economics and computer science from the Moscow Institute of Physics and Technology (1993), a PhD in applied math from the Russian Academy of Sciences (1994), and a habilitation (postdoctoral) degree in economics from the Academy (2002). He has had visiting positions at MIT and Princeton in the United States.
He was a rising star in Russia, and rector of the New Economic School in Moscow from 2004 to 2013, before he fled the country for France. Outspoken criticism led to Guriev falling foul of the government, which prompted interrogations and fears for his safety, as elaborated in The New York Times here and here.
He found a new academic home at Sciences Po in 2013. From 2016 to 2019 he took leave to serve as chief economist at the European Bank for Reconstruction and Development. From 2019 until his appointment as provost he was scientific director of Sciences Po’s masters and PhD programmes in economics.
AI in universities
There are many ways for AI, a broad term that covers several fields, to be used in universities. AI may help to advance education, where there are plenty of functions that are automatable, while research helps to develop AI.
“Basically, our job is not to make students read books, remember things, and take an exam. Our job is to enable students to live a successful and impactful life, which means we need to teach them how to use AI tools and how AI works. Students should also be able to think critically about what these tools can and cannot deliver,” said Guriev.
“You need a bit of computer science, a bit of math, and that can give you at least an idea where ChatGPT’s answers come from. On the other hand, AI, or more generally computer science, can help to eliminate things that are boring, and that you do not need to know.”
At Sciences Po, sometimes students are taught how the software they use works, and sometimes they are taught to write software. “But overall, once they understand the principle, students just need to use the programs, as that’s what they will do in the workplace.”
A major dimension of new technologies for universities is online learning. Using big data helps to deliver online learning effectively, enabling students to learn anywhere and in their own time. “I love teaching in person, but it is also great that many things can be taught asynchronously.
“Some people have mobility issues, some people live in different countries. In that sense, digital tools and AI can be an inclusion mechanism, when well thought through. We should be open to innovation because there are many demands on us to use technology in education.”
Rules and ethics and AI
Guriev told University World News that understandings of new technologies in social sciences eventually lead to how new technologies are regulated. “Europe is progressing quickly on regulating social media and digital platforms.”
The issue of controls over new AI technologies is currently in the public eye, following an open letter by many leading technology innovators and professors from around the world, titled “Pause giant AI experiments”.
The call was published on 22 March by the Future of Life Institute, a non-profit organisation that seeks to mitigate the existential risks posed by transformative technologies including AI. “We call on all AI labs to immediately pause for at least six months the training of AI systems more powerful than GPT-4,” said the letter, which spurred a flurry of responses in academia, for and against.
Last year Sciences Po, along with Georgetown University in the US, started working with a newly created Paris-based McCourt Institute, which supports research into social media and AI and aims to ensure that digital governance is prioritised in the development of new technologies and the web.
Regarding the need to teach ethics and AI, Guriev used the example of the Enron accounting scandal of 2001 and the global financial crisis in 2008, after which business schools the world over woke up to the need to teach ethics. “Greed is good, but it is not good enough,” he said.
“The same applies to AI. When we train future leaders, they should understand how AI works because they need to think about what is fair and unfair and how to change the rules of the game.
“Remember, the business leader’s job is to be successful in business, given the rules of the game.” The rules can be formal, such as government regulations, or informal, such as reputational considerations and societal attitudes, norms and beliefs. “Sometimes public opinion leaders are more important than lawmakers for success in business or in government.
“So in this sense, we need to teach people whether something is ethical or not. We teach people how we use reputational and regulatory mechanisms to limit non-ethical behaviour. And we tell students that when you make decisions, you have to think about the long term implications, because there is great hope that a future society will be more just.
“If you make an unjust decision today, it will backfire tomorrow,” said Guriev. “It is important to make sure that every student, every future leader, remembers this.”
An interdisciplinary approach
The confluence of sciences and social sciences entwines increasingly into the overarching approach of Sciences Po, along with forging partnerships with other institutions. The grande école has transformed into an international research university.
Sciences Po President Mathias Vicherat identified digital and environmental transformation as two of the most important challenges of today, echoing the European Union’s priorities. “For us, these are transversal interdisciplinary challenges, and we decided to make this mainstream,” said Guriev.
In January it launched the Transforming Interdisciplinary Education and Research for Evolving Democracies – TIERED – project. Sciences Po obtained funding from the French government and partnered with 10 institutions that work in disciplines such as the natural sciences, the digital sphere and marine biology.
The project enables all of the universities involved to draw on each other’s variety of disciplines. “We can work with people from other fields, and also bring them in to teach our students. That is a huge innovation. It is a big journey of 10 years that starts now,” he said.
Guriev embodies interdisciplinarity – as an economist, physicist and mathematician heading up the academic programme at a social sciences university – and believes in its importance.
“That’s why I do social science. I think the big challenges of our times are very much related to social sciences. Because you can have excellent technology, but if it is not used well, you can have a major problem. Both climate and digital are examples of this.”
It was decided every student should have at least one course on the environment, to mainstream that challenge. Sciences Po is a geographically distributed university with seven campuses across France and 15,000 students – half of them international students, representing 150 nationalities.
Kicking off in January, the university held an intensive 24-hour course over one week for an incoming cohort of 2,000 students. “That’s a lot of teaching. It is so students understand the scale of the problem, its social and political implications, the solutions and indeed social and political challenges related to those solutions.”
It is in the DNA of Sciences Po, Guriev said, to work on public policy. “So whenever we teach something, whenever we do research, we always think about policy implications. There is a huge potential for policies and policy solutions for the digital and environmental transformations.
“In this sense, it is great that we involve people from all fields, so students can actually see not just enormous and insurmountable problems, but also potential solutions that social sciences and natural sciences working together can provide.”