Author: Philip Garnett
Human-Computer Decision Making
The potential impact of advanced analytics, algorithms, and perhaps most importantly artificial intelligence (AI) on future employment is currently receiving a lot of attention. Particularly in the case of AIs replacing large numbers of low skilled jobs. However, the impact of advanced analytics, algorithms, and AI will not be limited to low skilled jobs. In fact it could be argued that many jobs have been feeling the impact of advanced analytics and AI for some time, and the changes the future will bring to all areas of employment will likely be significant.
Robert Harris in his novel, The Fear Index, paints a dystopian future where a rogue AI manipulates and kills employees of an advanced quantitative hedge fund, ultimately taking over the company and declaring that “in the future all companies will be alive”. Perhaps an extreme imagined future, but it does raise interesting questions about how we are going to work with our digital colleagues of the future. An equally dystopian alternative is of the inscrutable decision making machine that is blindly followed by the obedient human employee, “computer says no” (Carol Beer of BBC’s Little Britain).
The paper will explore the issues around working with computers in human-computer decision making processes, particularly their impact on front-line staff and middle management. Where the interaction between the human and computer is less of a (celebrated?) collaboration between domain expert and AI, and more of a sequence of steps where the human and AI respond to the inputs and outputs of the other. This situation presents a challenge for both human employee and manager. The employee cannot seek explanation for the decision of the AI, and therefore must choose whether nor not the recommendation is followed. What consequence does this have for both the role of discretion in decision making, and the situation of responsibility? Similarly, how does one manage an AI? A manager is also unable to seek the bases of an algorithmic decision, and therefore if the responsibility ultimately lies with a human how do we consistently determine the best management response to when the AI is correctly or incorrectly ignored? The paper will also explore the future of management itself, as the proposal of some quantitative hedge funds to replace managers with AIs suggests that the role of manager is also not beyond the reach of advance AI.