When faced with a decision regarding probability or heuristics, people generally show their bias toward a heuristic, even if it might be the wrong decision, such as on the classic base-rate neglect task (Kahneman & Tversky, 1973). The crucial question is whether people know that they are focused on this bias. Recent dual process theories (DPTs) have incorporated the crucial role of conflict detection and resolution to better explain why people are biased on classic reasoning and judgment tasks. Two recent models, the Logical Intuition Model (De Neys, 2012) and the Three-stages Model (Pennycook, Fugelsang, & Koehler, 2015) suggest that the source of errors and bias can be explained in two distinct ways. The Logical Intuition Model postulates that people are generally efficient and routine conflict detectors and that errors are due to a failure to inhibit an initial, intuitive response.

The Three-stages Model claims that detection is imperfect and that the detection mechanism is the main source of errors because people do not recognize that they are making biased decisions. These claims were investigated in a series of three experiments. In Experiment 1, participants completed a modified base-rate neglect task. In Experiment 2, a conditional reasoning task was added to test whether the claims of the two base-rate neglect models would transfer to a qualitatively different task. In Experiment 3, participants were placed into four different groups where they were given one of two forms of false feedback, true feedback, or no feedback (control) in order to test whether feedback would interact with answer confidence and resultant intuitive or analytic processing correlates (such as response time).

Across all three experiments, the Three-stages Model's claim that monitoring/detection failures is the main source of bias on the base-rate neglect task was supported over the Logical Intuition Model's claim of inhibition failures as the major source of bias. Experiments 2 and 3 support the explanation that these two models are task-specific to base-rate neglect, as conditional reasoning behavioral patterns did not support either model fully. Feedback did not have the predicted effects on accuracy, response times, or confidence. Implications of these findings regarding general dual process theory, including the impact of methodological limitations, are discussed. Small modifications to the Three-stages Model are offered to reflect the data presented here.