14th March 2019
Whoever has started to be supervised by an algorithm recognizes that at first, they feel odd, but over time they get used to it. In a fairly constrained environment, an AI-based system can make accurate decisions. The rest of a supervisor’s activities are trivial for the system:
- Assign tasks to team members by their availability and their skills profile.
- Verify compliance with these instructions.
- Assess effectiveness of these actions for meeting the proposed objectives. This assessment feeds back into the system, enabling the algorithm to make decisions with a higher success rate in the future.
Certainly, a boss of this kind does not bring about a high degree of bonding on a personal level, but how many human supervisors do? Aside from this, the benefits of an AI-based management system include:
- A decision-making process based on data, even if high volumes of information must be handled.
- Workload distribution through objective criteria, without preferences or exclusions.
- Equally objective performance appraisal.
This focus drastically breaks away from how we understood the role of technology in working environments over the last two centuries. Since the first industrial revolution, machines adopted a subordinate role as a toolkit: they were something we used to perform our work more efficiently, sticking strictly to our instructions. Now we are seeing a scenario in which certain tasks can be performed preferably by humans, but under supervision from artificial intelligence systems. Think of tasks to do with service, where we want to be served by a human being rather than by a robot: waiters in the hospitality industry, nursing staff in a hospital, and more. What stops an algorithm from doing tasks that allocate shifts, distribute les or beds, give instructions on dispensing food or drugs, collect data on customer or patient satisfaction, etc.?
The problem arises when we analyse, instead of efficiency, the ethical issues associated with this change in roles. Some experts have underscored the risks of this change. Lasse Rouhiainen, the author of “Artificial Intelligence: 101 Things You Must Know Today about Our Future”, upon analysing the initiatives in some Asian countries, commented: “My distrust is that maybe they follow the same ideology as China, that is, that they first develop applications and then review ethical issues, when they should be upside down.”
Among the ethical paradoxes of this situation, the IESE Business School Insight (Fall 2018) points out the following:
The debate on fairness.
We said earlier that supposedly an automated system will make decisions without prejudices nor subjective conditions. However, initial experiences show that decisions made by algorithms replicate some of the biases present in human decision-makers, since we are ultimately the ones who developed them. It is possible non-human leaders assign tasks that bring more added value or greater visibility to the company, preferably to men (than women) or to people of a certain ethnic group. If they use historical performance indicators contaminated by our current social structures in which men on average are more available than women, such decision patterns will be perpetuated.
What is characteristic of a boss is not only his ability to make decisions, but also to take responsibility for these decisions. If an automated system fails and gives us instructions that lead to undesirable consequences, who is responsible: the manufacturer, the implementer?
A boss should not constantly explain the reasons behind a decision of his, but he should be able to do so if justifiably asked to do so. The complexity intrinsic to artificial intelligence algorithms can lead to their decisions not being understood even by their developers. It is uncomfortable to work for a boss with whom you cannot negotiate any of the instructions he gives us, and without even understanding the reasons why he relays these commands.
It is not bad to have non-human bosses, so long as the ethical framework in which they act is carefully designed and supervised by human beings.