Jrcb4 The Impact of Artificial Intelligence on Learning final

2.4.1
 
Skills in economic studies of AI impact 
Much of the current economic research on the future of work and the impact of AI starts 
from analysing the impact of computers on skill demand. It is, therefore, important to 
understand how skills and work tasks have been interpreted in these studies. Below, we 
put these econometric studies in the context of the three-level model presented above 
41
These include, for example, Autor, Levy and Murnane (2003), Acemoglu and Restrepo (2016), and, in a 
more pessimistic vein, Brynjolfsson and McAfee (2012). Autor, in particular, has argued that the main 
impact of automation has been in the polarization of labour markets. He also argues that the use of AI will 
increase the comparative advantage of humans in tasks that require problem-solving skills, adaptability, 
creativity, flexibility, and common sense (Autor 2015). A recent collection of articles on the economy of AI 
is available from the US National Bureau of Economics Research (Agrawal, Gans, and Goldfarb 2018). Many 
of these studies, however, could be put in a somewhat different light by looking time use and hours worked 
in the economy per capita. For example, in Finland the time used for paid labour has decreased about one 
fifth per capita in the last forty years. 
42
The global and networked character of knowledge economy poses some quite deep methodological 
challenges here. We have extensive economic data on the national level, and it is therefore natural to 
assume that we should use those data as a starting point to study the economic impact of computerization 
and AI. The available data, however, do not necessarily capture the non-local and functional aspects of 
economy. In biology, the observation that those aspects of living systems that make them “alive” cannot 
be described using data on their constituent components led in the 1950s to “relational biology.” It focuses 
on the functional organization of biological systems instead of their various material implementations 
(Rosen 1958, 1991; Rashevsky 1954, 1972). In particular, Robert Rosen argued that dynamic models, 
such as those used in physics and economics, are not able to capture the essence of biology as systems are 
alive because of complex networks of interrelated functions. A category theoretic formalism is needed to 
model such systems (cf. Louie 2009). 

---

19 
(see 2.1), showing that different types of AI have capabilities on different levels of this 
model. 
Many of the influential econometric studies use the U.S. Occupational Information 
Network (O*NET) database as a starting point.
43
O*NET contains now about 1000 
occupational definitions to help students, job seekers, and educators to understand skill 
requirements and work content in different occupations. An example of the task structure 
of one occupation, “Middle School Teachers, Except Special and Career/Technical 
Education,” is shown in Figure 4.