In recent years, digital technologies shaped all aspects of the current socio-economic scenario. The relation between these new technologies and workers is a classical controversy. If on one hand digitalization allows firms to substitute tasks previously performed by workers, it is doubtless that the use of these digital technologies increases labor productivity and consequently impacts employment. A mismatch exists between skill demand and supply due to the complexity of the problem: technologies are very different from each other, and so are their impact on occupations. In this scenario, understanding and predicting which are the competences impacted by digital technologies is fundamental for preparing workers, firms and policy makers to address this digital wave. The study of the evolution of skills requirements in the labor market is well-established in the literature [1] and initially such phenomenon was mainly linked to routine based activities that can easily be performed by sophisticated algorithms. Recently, also the automation of tasks which have always been considered too complex to be performed by a technology seems like a plausible scenario [2]. Moreover, soft skills are ever more recognized as a bottleneck for computerisation, since machines cannot replicate what is uncodable [3]. Frey and Osborne [4] quantitatively estimate the computerisation susceptibility of job profiles on 702 detailed occupations collected in the O*NET database (https://www.onetonline.org/). Several authors have investigated the results of Frey and Osborne [4] in the past years [5] or propose new methodologies to study the topic [6]. Despite the large literature that exists on the subject, there is still a lack in quantitative measures of the effect that automation will have on what workers do. The most similar study in this direction has been done by Brandes and Wattenhofer [5], that refine Frey & Osborne’s results assigning automation probabilities to tasks. Anyway, our study goes deeper in this direction studying how the risk of automation is linked to skills, abilities and knowledge. Like other authors who deepened Frey & Osborne [4] results, we use their output to switch the focus, from jobs to competences. To fully understand our method it is important to describe how O*NET is structured. O*NET taxonomy, developed by the U.S. Department of Labor, contains abilities, skills and domains of knowledge to perform a job. O*NET distinguishes the competences in 3 macro-categories: (1) abilities that are enduring attributes of the individual that influence performance; (2) skills that are developed capacities that facilitate learning or the more rapid acquisition of knowledge; (3) knowledge that is an organized set of principles and facts applying in general domains. We use the term “competence” to refer either to abilities, skills or knowledge. Each job profile has quantitative information about “importance” and “level” for every owned competence. The “importance” answers the question “How important is a given competence to the performance of a given job?” whereas the “level” answers “What level of a given competence is needed to perform a given job?”. We estimate the probability of computerisation of each competence using the computerisation probability of the occupations and the “importance” and “level” information. Using this approach, we are able to give statistical evidence of different levels of automation probability of different competence groups. We found that the computerisation probability of the macro-class “ability” is 0.50, that is greater than those of “skills” (0.40) and “knowledge” (0.36). This is a reasonable result since the abilities comprehend enduring factors of workers (such as speed of limb movement, control precision, rate control etc.) that are more simple to codify in a sophisticated algorithm than complex concepts to learn and use at the right time (i.e. knowledge) or soft and technical skills, that could be acquired through experience. Among the abilities those at greatest risk of automation are reaction time/speed abilities (0.61). Differently, idea generation (as an ability) has a lower probability of automation, 0.39. The skills more susceptible to automation are the technical skills, such as equipment maintenance (0.63). In contrast, the systems skills i.e. developed capacities used to understand and improve socio-technical systems, have a low level of computerisation (0.35). Finally, among knowledge we could distinguish high automation risk knowledge, such as therapy/counseling (0.20), and knowledge with a low computerisation probability, as for example mechanics (0.53). From an academic point of view, we offer a quantitative approach to measure to what extent competences are minimizing the risk of being substituted by machines, giving statistical evidence of different levels of automation probability of different groups of competence. Moreover, we offer a holistic view: new technologies bring new opportunities (e.g. automation) but also new needs (e.g. managing automation). Most of the state of the art deals with the aggregate employment impact of innovation, and does not disentangle the analysis in terms of competences. Finally, our results can help companies and policy makers to estimate the impact of automation on competences. Given that there is less debate about the positive employment effect of innovation, a quantitative understanding of this phenomena, possibly free from negative or positive biases, can help face the future of training and hiring. References [1] Cedefop, (2019). Online job vacancies and skills analysis: a Cedefop pan-European approach. [2] Colombo, E., Mercorio, F., & Mezzanzanica, M. (2019). AI meets labor market: exploring the link between automation and skills. Information Economics and Policy. [3] Acemoglu, D., & Autor, D., (2010). Skills, tasks and technologies: implications for employment and earnings. doi:10.3386/w16082 [4] Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation?. Technological forecasting and social change, 114, 254-280. [5] Brandes, P., & Wattenhofer, R. (2016). Opening the Frey/Osborne black box: Which tasks of a job are susceptible to computerization?. arXiv preprint arXiv:1604.08823. [6] Montobbio, F., Staccioli, J., Virgillito, M. E., & Vivarelli, M. (2022). Robots and the origin of their labour-saving impact. Technological Forecasting and Social Change, 174, 121122.

Here Comes the Robot: Measuring the Risk of Automation of Human Competences through a Quantitative Approach

Vito Giordano
Primo
;
Gualtiero Fantoni
2022-01-01

Abstract

In recent years, digital technologies shaped all aspects of the current socio-economic scenario. The relation between these new technologies and workers is a classical controversy. If on one hand digitalization allows firms to substitute tasks previously performed by workers, it is doubtless that the use of these digital technologies increases labor productivity and consequently impacts employment. A mismatch exists between skill demand and supply due to the complexity of the problem: technologies are very different from each other, and so are their impact on occupations. In this scenario, understanding and predicting which are the competences impacted by digital technologies is fundamental for preparing workers, firms and policy makers to address this digital wave. The study of the evolution of skills requirements in the labor market is well-established in the literature [1] and initially such phenomenon was mainly linked to routine based activities that can easily be performed by sophisticated algorithms. Recently, also the automation of tasks which have always been considered too complex to be performed by a technology seems like a plausible scenario [2]. Moreover, soft skills are ever more recognized as a bottleneck for computerisation, since machines cannot replicate what is uncodable [3]. Frey and Osborne [4] quantitatively estimate the computerisation susceptibility of job profiles on 702 detailed occupations collected in the O*NET database (https://www.onetonline.org/). Several authors have investigated the results of Frey and Osborne [4] in the past years [5] or propose new methodologies to study the topic [6]. Despite the large literature that exists on the subject, there is still a lack in quantitative measures of the effect that automation will have on what workers do. The most similar study in this direction has been done by Brandes and Wattenhofer [5], that refine Frey & Osborne’s results assigning automation probabilities to tasks. Anyway, our study goes deeper in this direction studying how the risk of automation is linked to skills, abilities and knowledge. Like other authors who deepened Frey & Osborne [4] results, we use their output to switch the focus, from jobs to competences. To fully understand our method it is important to describe how O*NET is structured. O*NET taxonomy, developed by the U.S. Department of Labor, contains abilities, skills and domains of knowledge to perform a job. O*NET distinguishes the competences in 3 macro-categories: (1) abilities that are enduring attributes of the individual that influence performance; (2) skills that are developed capacities that facilitate learning or the more rapid acquisition of knowledge; (3) knowledge that is an organized set of principles and facts applying in general domains. We use the term “competence” to refer either to abilities, skills or knowledge. Each job profile has quantitative information about “importance” and “level” for every owned competence. The “importance” answers the question “How important is a given competence to the performance of a given job?” whereas the “level” answers “What level of a given competence is needed to perform a given job?”. We estimate the probability of computerisation of each competence using the computerisation probability of the occupations and the “importance” and “level” information. Using this approach, we are able to give statistical evidence of different levels of automation probability of different competence groups. We found that the computerisation probability of the macro-class “ability” is 0.50, that is greater than those of “skills” (0.40) and “knowledge” (0.36). This is a reasonable result since the abilities comprehend enduring factors of workers (such as speed of limb movement, control precision, rate control etc.) that are more simple to codify in a sophisticated algorithm than complex concepts to learn and use at the right time (i.e. knowledge) or soft and technical skills, that could be acquired through experience. Among the abilities those at greatest risk of automation are reaction time/speed abilities (0.61). Differently, idea generation (as an ability) has a lower probability of automation, 0.39. The skills more susceptible to automation are the technical skills, such as equipment maintenance (0.63). In contrast, the systems skills i.e. developed capacities used to understand and improve socio-technical systems, have a low level of computerisation (0.35). Finally, among knowledge we could distinguish high automation risk knowledge, such as therapy/counseling (0.20), and knowledge with a low computerisation probability, as for example mechanics (0.53). From an academic point of view, we offer a quantitative approach to measure to what extent competences are minimizing the risk of being substituted by machines, giving statistical evidence of different levels of automation probability of different groups of competence. Moreover, we offer a holistic view: new technologies bring new opportunities (e.g. automation) but also new needs (e.g. managing automation). Most of the state of the art deals with the aggregate employment impact of innovation, and does not disentangle the analysis in terms of competences. Finally, our results can help companies and policy makers to estimate the impact of automation on competences. Given that there is less debate about the positive employment effect of innovation, a quantitative understanding of this phenomena, possibly free from negative or positive biases, can help face the future of training and hiring. References [1] Cedefop, (2019). Online job vacancies and skills analysis: a Cedefop pan-European approach. [2] Colombo, E., Mercorio, F., & Mezzanzanica, M. (2019). AI meets labor market: exploring the link between automation and skills. Information Economics and Policy. [3] Acemoglu, D., & Autor, D., (2010). Skills, tasks and technologies: implications for employment and earnings. doi:10.3386/w16082 [4] Frey, C. B., & Osborne, M. A. (2017). The future of employment: How susceptible are jobs to computerisation?. Technological forecasting and social change, 114, 254-280. [5] Brandes, P., & Wattenhofer, R. (2016). Opening the Frey/Osborne black box: Which tasks of a job are susceptible to computerization?. arXiv preprint arXiv:1604.08823. [6] Montobbio, F., Staccioli, J., Virgillito, M. E., & Vivarelli, M. (2022). Robots and the origin of their labour-saving impact. Technological Forecasting and Social Change, 174, 121122.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11568/1138564
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact