The algorithm, a tool against populist rhetoric

There is no shortage of critiques concerning the use of AI in public decision-making processes. Scholars, for example, have described algorithms as a “toxic cocktail for democracy” by pointing at the ever-growing availability of big data that can be manipulated to influence public discourse in specific (problematic) directions. Can we flip over the argument and make a case for harnessing the potential of digital technologies to increase the quality of democracy in times of rampant populism? We can actually resort to democratic theories that build on purely input frames to encompass the complex dimension of both political processes and policymaking.

Let us attempt to provide a non-exhaustive classification by looking at the steps of policymaking and decision-making procedures. In the context of public decision-making, algorithms can function in several ways. AI can: a) represent the world by, for example, proxying demographic data; b) predict or test the desirability of a given course of action in light of the results that the latter may determine, such as immigration detention risk assessment; c) reach a decision, generally on the assumption of the inherent reliability of the result and/or efficiency of the process, by, for example selecting individuals who will benefit from an allocation decision that had been made within the traditional political process; and d) act as an algorithm-manager by supervising and controlling public servants who are required to make complex decisions.

In each of these scenarios, algorithms play different roles. Representation and prediction functions imply that algorithms can provide decisionmakers with accurate information concerning a given political option. Therefore, algorithms do not replace political choices, but instead they create the conditions for a political alternative to be confronted with concrete outputs. The selection function uses algorithms to speed up procedures that would otherwise require careful and lengthy examination. Algorithms can thus ensure the efficiency of the selective process and the consistency of results. In the scenarios depicted above under c) and d), the algorithm makes decisions instead of the decisionmaker identified via the ordinary political process.

Against this backdrop, algorithmic decision-making is unfit to any of those tasks if it is expected to reach substantive decisions independently. In contrast, AI can boost democratic processes if some conditions for its proper functioning are observed. In particular, by unpacking the logic of democratic legitimacy, we can conclude that AI needs to be coupled with the need for 1) understanding and selecting civic issues that deserve to be addressed by political institutions; 2) controlling which issues reach the democratic institutions; 3) evaluating and challenging the results of a given course of a decision taken by or on account of an AI. In other words, algorithms can increase democratic legitimacy at times of rampant populism, provided that their use takes place within a framework that maximizes political equality and rational decision-making by enabling wider participation, consideration of diverse social issues and oversight of the decisions made.

Under those conditions, the efficiency and the representation of reality that an algorithmic process may produce expose the fallacies or the “easy truth” of populist rhetoric. Algorithmic decision-making is not good because the output is intrinsically trustworthy. It is good as long as it is embedded in a democratic frame that enables both the represented and representatives to exercise choices and control over the decision-making process. In this way, algorithms can expose the populist rhetoric by being an instrument of knowledge and, therefore, a tool to read reality and help solve its problems.

Provided by
Bocconi University

There is no shortage of critiques concerning the use of AI in public decision-making processes. Scholars, for example, have described algorithms as a “toxic cocktail for democracy” by pointing at the ever-growing availability of big data that can be manipulated to influence public discourse in specific (problematic) directions. Can we flip over the argument and make a case for harnessing the potential of digital technologies to increase the quality of democracy in times of rampant populism? We can actually resort to democratic theories that build on purely input frames to encompass the complex dimension of both political processes and policymaking.

Let us attempt to provide a non-exhaustive classification by looking at the steps of policymaking and decision-making procedures. In the context of public decision-making, algorithms can function in several ways. AI can: a) represent the world by, for example, proxying demographic data; b) predict or test the desirability of a given course of action in light of the results that the latter may determine, such as immigration detention risk assessment; c) reach a decision, generally on the assumption of the inherent reliability of the result and/or efficiency of the process, by, for example selecting individuals who will benefit from an allocation decision that had been made within the traditional political process; and d) act as an algorithm-manager by supervising and controlling public servants who are required to make complex decisions.

In each of these scenarios, algorithms play different roles. Representation and prediction functions imply that algorithms can provide decisionmakers with accurate information concerning a given political option. Therefore, algorithms do not replace political choices, but instead they create the conditions for a political alternative to be confronted with concrete outputs. The selection function uses algorithms to speed up procedures that would otherwise require careful and lengthy examination. Algorithms can thus ensure the efficiency of the selective process and the consistency of results. In the scenarios depicted above under c) and d), the algorithm makes decisions instead of the decisionmaker identified via the ordinary political process.

Against this backdrop, algorithmic decision-making is unfit to any of those tasks if it is expected to reach substantive decisions independently. In contrast, AI can boost democratic processes if some conditions for its proper functioning are observed. In particular, by unpacking the logic of democratic legitimacy, we can conclude that AI needs to be coupled with the need for 1) understanding and selecting civic issues that deserve to be addressed by political institutions; 2) controlling which issues reach the democratic institutions; 3) evaluating and challenging the results of a given course of a decision taken by or on account of an AI. In other words, algorithms can increase democratic legitimacy at times of rampant populism, provided that their use takes place within a framework that maximizes political equality and rational decision-making by enabling wider participation, consideration of diverse social issues and oversight of the decisions made.

Under those conditions, the efficiency and the representation of reality that an algorithmic process may produce expose the fallacies or the “easy truth” of populist rhetoric. Algorithmic decision-making is not good because the output is intrinsically trustworthy. It is good as long as it is embedded in a democratic frame that enables both the represented and representatives to exercise choices and control over the decision-making process. In this way, algorithms can expose the populist rhetoric by being an instrument of knowledge and, therefore, a tool to read reality and help solve its problems.

Provided by
Bocconi University