Choice Architecture: A Message and Environment Perspective

The thesis discusses a novel perspective for the analysis and design of a ‘choice architecture’, that is, the way in which a choice situation is ‘built’ and presented to the decision maker. The main theme is that a choice architecture can be conceptually divided into two dimensions: ‘Message’ and ‘Environment’. ‘Message’ refers to the description of options and information or suggestions about the choice at hand, such as: “Online subscription for $59”, “Most users choose X”, “Your peers drink less than you think”. ‘Environment’ refers to every element that the decision maker can perceive through her senses and interact with, such as a sensory cues, obstacles and constraint to decision, and the order or prominence of alternatives. It is argued that this distinction can be useful to inform policy making, specifically in the design of ‘nudges’—aspects of the choice architecture that work on rationality failures. It is also shown that this distinction can be applied for the conception of novel research questions and experiment designs to address them. The thesis is divided into three chapters. Chapter 1 reviews the literature on nudges and choice architecture, focusing on the discussion of three main issues related to nudging—definition, justification, and effectiveness—in the hope of ‘nudging’ future research to fill the gaps highlighted in the paper. Chapter 2, co-authored with Ivan Moscati and accepted for publication in Behavioural Public Policy, presents the Message-Environment distinction and its applications for policy making, specifically the design and analysis of nudges and choice architectures. Chapter 3 tests experimentally two framings at the Message and Environment dimensions of a risk-aversion elicitation context—specifically, multiple price list in the version of Holt and Laury (2002). According to the treatment, it was applied either a Message framing—where probabilities were expressed either as percentage (e.g., “10%”) or ratios (e.g., “1 out of 10”)—and/or an Environment framing—where the pie chart used to represent probabilities was either “standard” (i.e., having one slice for each outcome) or “sliced” (i.e., subdivided in ten equal slices). It is argued that, although the Message and Environment framings are somehow consistent with one another, their effect on risk aversion differs.