Unraveling the Algorithms for Humanized Digital Work Oriented Artificial Intelligence

For a better understanding, the author brings some examples of improvements made possible by the introduction of AI technologies on the factory floor, let us see:

[…] improved quality control (e.g. turbofan blades can be 3D inspected by the manufacturer with micrometer accuracy), made predictive maintenance possible (e.g. failure can be stopped in equipment by detecting even subtle changes with the help of networked sensors), reduced energy and material costs, inventory optimization, product design (e.g. aircraft parts were created by airbus that are much lighter than those designed by humans), improved safety and environmental performance.

On this matter, the Proposal for regulation of the European Parliament and of the Council laying down harmonized rules on artificial intelligence (Artificial Intelligence Act) and amending certain union legislative acts, by the European Comission, of april 2021, is surgical. This is because it provides for a set of horizontal mandatory requirements that must be observed in the various systems equipped with AI, with the scope of limiting its use for purposes beneficial to society, based on ethical principles, enunciating mechanisms that endow automated decisions with transparency and of traceability to ensure security and protection of fundamental rights throughout the entire life cycle of AI systems. The “AI Act”, in addition to being multidirectional and reaching several sensitive points of confluence of the use of these technologies in the different fields of application, guides them to the collective interest, addressing specific issues such as the necessary solutions for the effectiveness of regulation in the domain of employment, worker management and access to self-employment, lists high risk points for the misuse of algorithms in recruitment and selection processes, decision-making on promotions and dismissals, division of tasks and control or evaluation of people within the scope of contractual employment relationships, especially in terms of inspection and penalties arising from practices prohibited by law (e.g. in AI mechanisms that influence people or benefit from the vulnerabilities of specific groups of people due to age, disability, technical ignorance, causing physical or psychological damage to them) or the execution of high-risk AI systems that are not structured in accordance with the training, validation and testing data provided with the quality criteria established by law, which enable the governance and management of data externally. By conferring a high degree of risk to issues related to the algorithms applied to work, the proposal collaborates with the protection of the rights of workers of digital platforms, as it is concerned with the impacts that the misuse of AI generates on the career and in the livelihoods of these people.

Free translation by the author.

Because AI targets the realization of human activities by machines. Therefore, it is analyzed in parallel with biological and cognitive activities. As clarify De Christo Hundertmarck & Weber [7], the machine equipped with AI is capable of learning, and this machine learning is derived from two strands, machine learning, “which are Artificial Intelligence systems, subjective, able to learn from their own experience, endowed with ‘behavioral self-sufficiency’”; and machine deep learning, which uses “cytoartificial” neural networks (simplified simulations of how biological neurons behave) and extracts patterns and rules of human behavior via datasets, transcribing organic biological functions to machines in hardware format.

The Judgment of the Supreme Court of Spain ruled in a pioneering way in relation to the issue of algorithmic discrimination, in Judgment n. 805/2020, of September 25, 2020, in Appeal no. 4746/2019, of the Sala de lo Social Pleno for the unification of doctrine, in a lawsuit against the Glovo platform, in which it was declared that the relationship between the riders and the former is of a work-related nature. The decision highlights the functioning of the platform's algorithmic service management as a “subtle” way of exercising corporate power in relation to the way in which the service is provided by controlling its execution in real time (constant geolocation). This is how the domain of prices, of workers’ evaluations, of the distribution of tasks is exercised. In the judgment of the “Tribunale Civile di Milano” (Sezione specializzata in materia di impresa Ricorso ex art. 840 sexiesdecies c.p.c.) against Deliveroo Italy srl, it is mentioned that the algorithm manages the complex planning system, the acquisition of the order by the user, identification of passengers receiving the delivery proposal, the distribution and organization of workflows, in addition to monitoring the different stages of delivery, receipts and payments. During this management, the judgment showed that the algorithm proceeded in a discriminatory way, as it started distributing a smaller amount of tasks as a form of sanction to pilots who “did not respect the work model imposed on them or engaged in non-conforming conduct through two appropriate parameters of ‘reliability’ and ‘participation’” (n. 79). Thus, they were lowered in their scores by the algorithms and, as a result, began to receive fewer demands for work.