Preference, Belief, and Similarity: Selected Writings

The critical difference is curatorial authorship. Tversky selected these forty articles himself in collaboration with editor Eldar Shafir during the last months of his life, making the volume his own retrospective statement about which work mattered most. The 1982 Kahneman, Slovic, and Tversky edited collection focused specifically on the heuristics and biases program, whereas this volume spans Tversky's entire career across similarity theory, probability judgment, and preference modeling. The inclusion of articles on the 'hot hand' in basketball and arthritis pain beliefs alongside prospect theory demonstrates Tversky's own sense of his intellectual range, not an editor's posthumous guess at what belonged together.

Tversky's feature-based contrast model, presented in the 1977 'Features of Similarity,' rejected the assumption that similarity can be represented as distance in a multidimensional space. Instead, he argued that similarity judgments depend on the weighting of common and distinctive features, and critically, that these weights are context-dependent and asymmetric. The similarity of A to B is not necessarily the same as B to A. This framework, developed across several papers in the volume's Similarity section, provided a more psychologically plausible account than spatial models and directly influenced later work on categorization, analogy, and prototype theory. The section includes collaborations with Shmuel Sattath, Itamar Gati, and J. Wesley Hutchinson that extend and test the model.

The Preference section is subdivided into three phases: probabilistic models of choice, choice under risk and uncertainty, and contingent preferences. This structure traces a trajectory from Tversky's early work on intransitivity and elimination-by-aspects, through the prospect theory papers with Kahneman that challenged expected utility theory, to later work on preference reversals, reason-based choice, and context-dependence in legal decision-making. The section demonstrates that Tversky did not simply critique rational choice models from outside; he engaged with their mathematical formalisms, as evidenced by papers like 'Advances in Prospect Theory: Cumulative Representation of Uncertainty' which proposes a technical alternative to expected utility rather than merely documenting anomalies.

The most sustained critique came from Gerd Gigerenzer and the ABC Research Group, who argued in works like 'Bounded Rationality: The Adaptive Toolbox' that heuristics should be understood as ecologically rational strategies rather than sources of systematic error. Gigerenzer contended that Tversky and Kahneman's experimental designs often stripped away environmental structure that makes heuristics effective in real-world settings. This debate, documented across multiple review articles and books, reshaped how researchers frame the 'biases' language. Within the volume itself, Tversky's papers on support theory and the conjunction fallacy provide the primary material that subsequent researchers have either extended or contested in this ongoing debate about ecological rationality versus error.

These papers demonstrate Tversky's commitment to testing cognitive principles in domains where statistical reasoning has practical consequences. The 'hot hand' papers with Thomas Gilovich examined whether basketball fans and players misperceive streakiness in shooting performance, finding that the belief in hot streaks is largely a cognitive illusion imposed on essentially random sequences. The arthritis pain study with Donald Redelmeier showed that patients' beliefs about weather-pain correlations were driven by selective memory rather than actual symptomatic patterns. These papers, sometimes overlooked by readers who focus on the theoretical pieces, exemplify Tversky's method of using vivid, culturally salient examples to demonstrate that intuitive statistical reasoning fails even when the stakes are personal and immediate.

Tversky died in June 1996. According to the volume's front matter, he had selected the articles in collaboration with Eldar Shafir, a former student, close friend, and collaborator, during the final months of his life. The book finally appeared through MIT Press in November 2003. Editorial choices included retaining original article formatting rather than rewriting for consistency, adding brief introductory remarks before each section, and including a complete chronological bibliography of Tversky's work. The delay between Tversky's death and publication reflects the scale of the project and the editorial labor required to assemble and contextualize four decades of technical papers. The resulting volume is therefore both an authored and edited work, with Shafir's mediating role shaping how Tversky's legacy is presented.