Three ways AI can substantially improve economic analyses

The recent boom in machine learning (ML) and artificial intelligence (AI) techniques has shown their potential to revolutionize our societies, with noticeable advances in applied sciences (automation of transport, health, and interpersonal interaction). However, these powerful tools still have relatively low popularity in the social sciences. How can AI and ML be useful to a social science researcher, especially an economist like myself who studies innovation? Let me outline three ways in which AI can substantially improve economic analyses. First, AI offers the possibility of analyzing new data sources. In my work as assistant professor in economics and faculty affiliate at the Schwartz Reisman Institute for Technology and Society at the University of Toronto, I am interested in different aspects of innovation: patentable innovation and non-patentable innovation. While it is undeniable that technological and scientific innovations play a pivotal role in guaranteeing sustainable growth, innovation remains somewhat abstract and difficult to quantify. How can we classify inventions, ideas, and knowledge production on a scale measuring their degree of innovation? To answer these questions, I use textual information—by definition, a type of unstructured data. More precisely, it is widely proven that patent level counts do not capture technological innovation's real trends. They omit the economic, scientific, and social value of the patent. Moreover, they could be an artifact of administrative or cyclical economic conditions. Patent citations counts are especially not a good indicator of innovation in Canada, where most citations are done by patent examiners, not patent applicants. Instead, capturing technological innovations by making use of the patent's topical content and extracting new concepts found in their descriptions proves to be a good alternative. Therefore, one key step for an economist to benefit from AI is rethinking the notion of data. Whereas traditional data involved using, say, surveys or accounting data, the word "data" takes on a broader dimension in our era. Data can be textual information in newspapers, banks and company reports, patent descriptions, scientific articles, images, etc. All of these can be transcribed into processable languages using a type of AI called deep learning. Second, AI facilitates the analysis of much more data than before (this is what the term ‘Big Data’ refers to). For example, in one of my projects with Vasia Panousi, assistant professor of economics at Université de Montreal, we gathered nearly 2.5 million patents, and for each, a document of more than 100 pages. As you can imagine, this requires significant computing power—no longer a limitation with AI. Third, we can improve our knowledge of inequality with AI. In my paper on "Innovative Ideas and Gender Inequality," I look at whether women received credit for their intellectual production by using machine learning methods. These powerful new tools enabled me to answer an equally tricky question. The findings? Academic papers omitted from references are 20 per cent more likely to be female-authored than male-authored, with all-male author teams omitting more female papers. Moreover, men received the same number of citations for a median quality paper as women with a top 25 per cent quality paper. I show that a history of omission like this is related to future adverse productivity outcomes. However, increased diversity at editorial boards is associated with reducing this bias. I got many comments from people who found my results surprising because they did not feel like they paid attention to authors' names before citing them. But this is where the novelty lies: in both the usefulness and subtlety of AI methods. Indeed, they may reveal (un)conscious bias. To conclude, AI is a boon for social sciences, particularly economics. For more conservative researchers, AI tools offer an excellent complementarity to econometrics—a branch of economics that relies heavily on mathematical and statistical analysis. They also increase approaches for creating new databases, forecasting, and analyzing latent phenomena. But to fully benefit from AI analyses of innovation across Canada, we must increase collaboration between government and universities. The patent analysis techniques described above, for example, can markedly improve the patent examination and granting process by identifying high potential patents ex-ante. Also, because machine learning relies on large quantities of data, there is an urgent need to facilitate data access for researchers—with, of course, total respect for ethics and privacy. Marlène Koffi is an assistant professor in economics at the University of Toronto and a faculty affiliate at the Schwartz Reisman Institute for Technology and Society. The Hill Times

Three ways AI can substantially improve economic analyses

Artificial intelligence is an innovative tool social science researchers can use to aid their work. Social scientists, especially economists, should understand the ways in which it can make the world less confusing.

Canada’s aging crisis and a call for reform in eldercare

As COP29 wraps up, will Canada finally stop funding fossils this fall?

Addressing the threat of antimicrobial resistance

The wisdom of Homer Simpson and Aaron Sorkin

Climate finance should target resilient food systems

Trump’s ‘America First’ plans may undermine global trade

‘The devil’s in the details, and we don’t have any’: critics, civil liberties groups decry feds’ lack of clarity on changes to privacy and AI bill

AI is not intelligent and needs regulation now

Can biotechnology put patients’ needs ahead of profits?

AI IQ: safeguarding elections requires more oversight to thwart voter manipulation, say expert, Senator

Public interest in AI offers ‘unique moment’ for regulatory action

News

Get free news updates

Opinion

Submit a guest opinion article

resources

Advertise

subscribe

get help

Get free news updates

Submit a guest opinion article

Three ways AI can substantially improve economic analyses

Artificial intelligence is an innovative tool social science researchers can use to aid their work. Social scientists, especially economists, should understand the ways in which it can make the world less confusing.

Get The Policy Briefings Newsletter