The Open University
Public Choice, Evidence And Values: An Exploration Of The Theory And Practice Of Government Decision-Making
Abstract
dc:description.abstractThis thesis examines how evidence is used in complex public choices that involve weighing up multiple values or <i>evaluative criteria</i>. My primary aim is to shed fresh light on an old question that is being asked with renewed urgency in the wake of recent advances in artificial intelligence: In a democracy, what is the role of technocracies and the algorithms that they use to determine the best option for society?<br></br><br></br>Three extended essays address the question. The first two focus on the normative limits to algorithmic public choice in a democracy, i.e., the rules that define which choices can be automated and meet some basic concept of democratic legitimacy. I critique the internal logic and claims underpinning one prevailing set of algorithms - benefit-cost analysis - which is that just about everything has a price and can be compared in monetary terms. The claim has proponents in mainstream economics and political and moral philosophy. If true, there is an algorithm that succinctly captures any set of competing human interests for any policy options and identifies a unique ranking of options from best to worst. However, under a set of conditions that I refer to as weak incommensurability (WI), no aggregation rule meets a basic democratic legitimacy test. If WI holds, using an algorithm to make decisive social choices over-steps reasonable normative limits. Such <i>algorithmic overreach</i> is a primary peril of AI. Essay I uses some mathematical demonstrations and propositional logic to define weak incommensurability and to characterise its implications.<br></br><br></br>Essay II is an empirical examination of the complex decision-making behind restrictions on two civil liberties during the COVID-19 pandemic in the US. I use a variety of daily state-level panel data sources to run two econometric models. The essay demonstrates that the valuation of civil liberties is possible, but also clearly illustrates why the choice of valuation is driven by political persuasion. No single “correct” price that reflects the views of the majority of voters or their representatives could be employed in evaluations. Essay II thus provides an empirical example of WI and algorithmic overreach. <br></br><br></br>If WI holds, and decisive choices are through public consultation and voting (or voting by elected representatives), what is the role of technocracy? Specifically, can technocratic institutions improve the effectiveness of public deliberative processes, against the basic political and epistemic metrics of success? <br></br><br></br>To address these questions, Essay III compares the effectiveness of a standalone one-shot deliberative institution (with no technocratic involvement) with a <i>deliberative system</i>. A deliberative system is a series of communicating institutions: two technocratic institutions prepare an evidence base that is then used by the terminal public deliberation and voting institution. I analyse the design implications of notions of democratic legitimacy, identifying a series of constraints that apply to one-shot institutions and deliberative systems. A simulation model compares the effectiveness of deliberative designs that meet these constraints for the two institutional forms. In general, the more stringent the legitimacy conditions, the stronger the argument is for the deliberative system with a prior evidence base that can be used to boost learning and correct biases in the electorate. The simulations broadly support the idea that provided technocratic institutions have their own strict legitimacy rules, they can boost epistemic and political goals and help to correct the known shortcomings of direct democracy. <br></br><br></br> In short, it is possible to identify hard limits to technocratic choice in the circumstances of WI. These circumstances are neither infrequent nor implausible. Indeed, they are probable in polarised settings. Even where these limits apply, there is still a role for technocratic institutions and algorithmic choice modelling in a democracy: to inform and frame public deliberation. However, the conclusion comes with the caveat that there is ample scope for technocratic manipulation of evidence, and the upstream epistemic institutions must abide by strict neutrality rules.
Degree
thesis:*- Name dc:type.qualificationname
- phd
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- The Open University
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Watkins, Benjamin
Rights
- Language dc:language
- en