In this project, we aim at developing a unifying theoretical framework for multiple paradigms all concerned with dependent numerical judgments. We will use computational modeling, Bayesian modeling, and empirical studies to develop this framework.

Dependent judgments are frequently investigated throught various paradigms. Anchoring describes the finding that irrelevant numerical values influence a subsequent judgment for an unrelated question. Advice taking is a paradigm in which participants receive advice after providing an independent judgment for a numerical judgment question. This advice is less than optimal incorporated into a subsequent judgments. Sequential collaboration is a paradigm in which participants are presented with judgment question and the answer of a previous participant before they decide whether to adjust or maintain the presented judgment. The hindsight bias describes the finding that remembering a previous judgment is biased towards the correct answer if this was presented before the memory task. Finally, seeding effects describe the finding that numerical judgments become more accurate if participants are presented with the correct answer to a related question.

These paradigms all describe how presented numerical values influence subsequent numerical judgments. However, they also differ in the judgment process such as including an initial judgment, allowing to opt out of providing a judgment, or the relation between the presented numerical value and the judgment question.

In this project, we establish a unifying theoretical framework for these dependent-judgment paradigms. In a first step, we extend an existing computational model of sequential collaboration to a computational model of dependent judgments that also includes the other four paradigms. Based on this model, we develop a Bayesian model that allows us to fit the computational model to empirical data. Lastly, we will conduct seven empirical studies to test predictions of the computational model concerning the influence of individuals' expertise, judgment extremity, initial judgments, opting out, and relevance of the presented numerical value.