Research Interests

     Over the last eight years, my research interests have focused on how people understand and utilize information to make judgments and decisions. Of particular interest are the mechanisms which underlie general abilities to reason through complex information when uncertainty is involved.  In these types of situations, the data needed to make a decision are often presented as complicated statistics which are notoriously difficult to understand. In my research, I employ a combination of quantitative and qualitative research methods and analyses to evaluate how people process statistical data, which has strong theoretical contributions for discerning how people may perceive and utilize statistics in reasoning and decision making. This research also has valuable practical implications as statistical reasoning is one of the foundational pillars required for scientific thinking. I plan to continue this research via several avenues in both theoretical and applied contexts.

Statistics and the Reference Class Problem

     It is easy to feel overwhelmed when presented with statistics, especially when the meaning of the statistical data is not clear. For example, what does it mean when the newscaster says there is a 20% chance of showers? Does that mean it will only rain 20% of the day? Or that only 20% of the area will get rain? Or that 20% of the possible rain will actually fall? Without a knowing the appropriate reference class, or group from which the data are drawn, reasoners are often forced to make a decision based on an improper assessment of the numbers provided. (The correct answer is that out of 100 days with these weather conditions, rain occurs on 20 of them.) Although this is a rather benign version of the reference class problem, difficulties with this issue extends well into the very core of understanding statistics.

     Statistical testing involves an inherently nested structure in which values are dependent on the expression of other values. Understanding these relationships are foundational for appropriate use and application of statistics in practice. However, difficulties understanding statistics has been widely documented throughout numerous fields, contributing to the current research crisis as well as patient diagnostic errors (e.g., Gelman & Loken, 2014; Ioannidis, 2005; Ioannidis, Munafò, Fusar-Poli, Nosek, & David, 2014; Pashler & Wagenmakers, 2012). Therefore, research that can improve general statistical literacy is highly sought after. 

     As a stepping stone toward the more difficulty reference classes in statistics, a slightly less complicated version of the reference class problem can be found in Bayesian reasoning tasks (e.g., Gigerenzer, Gaissmaier, Kurz-Milcke, Schwartz, & Woloshin, 2007; Gigerenzer & Hoffrage, 1995; Hoffrage, Krauss, Martignon, & Gigerenzer, 2015; Johnson & Tubau, 2015; Reyna & Brainerd, 2008; Sirota, Kostovičová, & Vallée-Tourangeau, 2015; Talboy & Schneider, 2017, 2018, in press). In these types of reasoning tasks, there are difficulties with representing the inherently nested structure of the problem in a way that clearly elucidates the correct reference class needed to determine the solution. Additionally, computational demands compound these representation difficulties, contributing to generally low levels of accuracy.

     In my own research, we have tackled the representational difficulties of reasoning by fundamentally altering how information is presented and which reference classes are elucidated in the problem structure (Talboy & Schneider, 2017, 2018, in press).  In a related line, we break down the computational difficulties into the component processes of identification, computation, and application of values from the problem to the solution (Talboy & Schneider, in progress). In doing so, we discovered a general bias in which reasoners tend to select values that are presented in the problem text as the answer even when computations are required (Talboy & Schneider, in press, in progress, 2018). Moving forward, I plan to apply the advances made in understanding how people work through the complicated nested structure of Bayesian reasoning tasks to the more difficult nested structure of statistical testing.

Reference Dependence in Reasoning

     While completing earlier work on a brief tutorial designed to increased understanding of these Bayesian reasoning problems through both representation and computation training (Talboy & Schneider, 2017), I realized that the reasoning task could be structurally reformed to focus on the information needed to solve the problem rather than using the traditional format which focuses on conflicting information that only serves to confuse the reasoner. In doing so, we inadvertently found a mechanism for reference dependence in Bayesian reasoning that was not previously documented (Talboy & Schneider, 2018, in press). 

     Reference dependence is the tendency to start cognitive deliberations from a given or indicated point of reference, and is considered to be one of the most ubiquitous findings through judgment and decision making literature (e.g., Dinner, Johnson, Goldstein, & Liu, 2011; Hájek, 2007; Lopes & Oden, 1999; Tversky & Kahneman, 1991). Although the majority of research documenting reference dependence comes from the choice literature, the importance of context in shaping behavior has also been noted in several other domains, including logical reasoning (Johnson-Laird, 2010), problem solving (Kotovsky & Simon, 1990), extensional reasoning (Fox & Levav, 2004)—and now in Bayesian reasoning as well (Talboy & Schneider, 2018, in press).

     I parlayed my previous research on representational and computational difficulties into the foundation for my dissertation, with an eye toward how reference dependence affects uninitiated reasoners’ abilities to overcome these obstacles (Talboy, dissertation). I also evaluated the general value selection bias to determine the circumstances in which uninitiated reasoners revert to selecting values from the problem rather than completing computations (Talboy & Schneider, in progress, in press). I plan to extend this line of research to further evaluate the extent to which a value selection bias is utilized in other types of reasoning tasks involving reference classes, such as relative versus absolute risk.

Advancing Health Literacy

     Although the majority of my research focuses on the theoretical underpinnings of cognitive processes involved in reasoning about inherently nested problem structures, I also have an applied line of research that focuses on applying what we learn from research to everyday life. We recently published a paper geared toward the medical community that takes what we learned about Bayesian reasoning and applies it to understanding the outcomes of medical diagnostic testing, and how patients would use that information to make future medical decisions (Talboy & Schneider, 2018). I also led an interdisciplinary team on a collaborative project to evaluate how younger and older adults evaluate pharmaceutical pamphlet information to determine which treatment to use (Talboy, Aylward, Lende, & Guttmann, 2016; Talboy & Guttmann, in progress). I plan to continue researching how information presented in medical contexts can be more clearly elucidated to improve individual health literacy, as well as general health decision making and reasoning.