Before the pandemic, Canadian firms were moving rapidly to make use of the latest data science techniques, artificial intelligence advances and robotic-related technologies. But the severe recession caused by lockdowns and other Covid-19 restrictions delayed many organizations’ investment and adoption plans for these new digital technologies, according to a study by two University of Toronto professors.
In their quest to track the impact of Covid-19 on the transition to the digital economy and the effects it will have on the labour force, Professor Kelly Lyons from the Faculty of Information and Economics Professor Michelle Alexopoulos had to overcome a lack of real time data by “getting creative” with both their data sources and methods. When they began their study in early 2021, most of the available research related to labour markets both pre-dated the onset of the pandemic and did not focus on the Canadian economy.
Their goal was to help fill these knowledge gaps. “If you’re worried about the transition for labour, we need to give you some accurate information on the pace of adoption,” said Lyons.
To accomplish this, she and Alexopoulos worked together with a team of student research assistants. They drew on a mix of text-based and non-text-based sources of data, and used a multi-faceted, mixed-methods approach to analyze data and synthesize findings.
Among other tactics, they scrutinized the official statistics that were available, recorded the occurrence of the term “data scientist” as a current job title on LinkedIn, looked at the number of searches and downloads for AI-related software, and reviewed think tank and consulting reports using so-called “distance reading” methods where computational methods are applied to pertinent texts.
The details on these and the many other methodologies they employed are fully described in their final report entitled Evaluating the Future of Skills, Jobs, and Policies for the Post COVID Digital Economy, which was released last month.
“Even the qualitative analyses became kind of quantitative in how we did it,” says Lyons, explaining that when the team read papers the old fashioned way – doing what’s called a close read – they would fill in a questionnaire at the end. The questions, formulated by researchers, required yes/no or multiple choice answers, and the results were analyzed by the team.
“I was surprised by the findings we were able to extract when we got creative with methods and analyzed both traditional and non-traditional sources of data,” says Alexopoulos.
When searching for patent filings related to AI and data science, for example, the researchers were hampered by the fact that detailed information doesn’t generally become publicly available until 18 months after filing. To assess the impact of Covid-19, they used a novel method, that involved collecting the weekly patent application data that was available for the period between January 2015 and June 2021 and then developed an analysis method that takes advantage of regular identified patterns pre-Covid to provide information about how patenting filings have likely been impacted by the onset of the pandemic.
Their estimates suggested growth rates for AI-related patents for the periods 2017-2018 and 2018-2019 pre-COVID were approximately 35%, with the growth rate for 2019-2020 falling to approximately 13.5%- 16.2%. Month over month comparisons also highlighted significant declines in AI-related applications occurring after the onset of the pandemic with some indication that the number of applications is now rebounding.
“The numbers are noisy but useful in terms of getting a real time index and a preliminary estimate of the pandemic’s impact,” said Alexopoulos.
She and Lyons hope that this, and the other data they uncovered in their report, will aid organizations and policy makers concerned about how the transition to digital technologies will affect the labour force.
While they concluded that the slowdown in commercialization and innovation likely means that the timing of future labour market disruptions and the possible emergence of skill gaps in many sectors of the economy will now occur a few years later than forecasts made prior to the pandemic suggested (with a few notable exceptions), the data revealed in the report also implies that, barring any unforeseen Covid impacts, employment opportunities related to AI and data science will quickly surpass their pre-pandemic levels and usher in increased productivity growth as the economy rebounds.
To spread the word about what they have uncovered and spark discussion about how to help smooth the transition to the digital economy and support displaced workers, the study’s authors are participating in several upcoming events including a seminar this week at the Schwartz Reisman Institute for Technology and Society and next month at the CASCON-Evoke 2021 conference for industry experts and academics in the computer science and software engineering fields.
You can also follow the Future Jobs Canada Twitter account and get updates on Lyons’ and Alexopoulos’ research on LinkedIn.
