1 How people do relational reasoning? Role of problem complexity and domain familiarity Shikhar Kumar School of Information: Science, Technology and Arts, University of Arizona, Tucson, AZ, Iliano Cervesato School of Computer Science, Carnegie Mellon University, Doha, Qatar Cleotilde Gonzalez Department of Social and Decision Sciences, Carnegie Mellon University, Pittsburgh, PA Abstract The goal of this paper is to study how people do relational reasoning, such as selecting the grade of all students in a class with GPA (Grade Point Average) greater than 3.5. Literature in the field of psychology of human reasoning offer little insight as to how people solve relational problems. We present two studies that look at human performance in relational problems that use basic relational operators. Our results present the first evidence towards the role of problem complexity on performance as determined by the accuracy and discrimination rates. We also look at the role of familiarity with tabular representation of information, as found in spreadsheets for example, and other factors for relational reasoning, and show that familiarity does not play a significant role in determining performance in relational problem solving, which we found counterintuitive. Keywords: Relational reasoning, Problem solving, Spreadsheet, Domain familiarity, Problem complexity 1. Introduction Nowadays, data are more easily accessible than ever, yet support for deriving interesting consequences from base data is often unavailable, too expensive, or too technical for many users. For example, a student may have access to prerequisite listings and expected offering dates of courses but have no way to sieve through possible course sequences unless the college provides a dedicated tool. Similarly, an investor may know the instruments held in his mutual fund portfolio but have no easy way to unravel them and reveal his exposure to a specific industry or company. In all cases, manually inferring useful information from raw data is time consuming and error prone, a situation that often results in bad decisions, suboptimal plans, or missed opportunities. In fact, there is currently no simple and general application that empowers users to compute useful inferences on raw data. Cervesato (2007; 2013) addressed this problem by drawing inspiration from a type of automated data inference that is immensely popular: the spreadsheet. Applications such as Microsoft Excel and others are readily available and allow users to routinely perform complex custom calculations on numerical data. The spreadsheet's clever interface makes it easy to use