PsycAssist

Few resources are available for the automatic generation of Raven-like matrices. Some of them (e.g., Sandia) are no longer working, while others (e.g., Corvus) are hardly customizable unless one has a medium-high expertise of programming in javascript languages. The ideal solution for the automatic generation of Raven-like matrices would be a software that can be easily maintained and updated in an accessible programming language. For this reason, a new software for the generation of Raven-like matrices (and relative distractors) is under development as an R package. This package is capable to generate Raven-like matrices following different rules concatenated with different directional logics (i.e., horizontal, vertical, diagonal). Differently from existing software (e.g., Corvus), this package is able to generate matrices considering different types of rules, starting from the most basic ones (e.g., changes in size, objects orientation) to the most complex ones, based on inferential and inductive reasoning. Having an open-source basis for such a software allows users to customize the generation of their own matrices according to new rules. Further information on the developmental stage of the package is available at https://ottaviae.github.io/stimuliGen/OfficialDocumentation.html

Raven's matrices are one of the most common measures for the assessment of fluid intelligence. All the variants of Raven’s matrices, namely the colored progressive matrices, the standard progressive matrices, and the advanced progressive matrices, can be designed according to some “generative rules”. These rules define the relationships between the elements represented in each cell of the matrix, and they can be combined according to different directional logics (i.e., horizontal, vertical, diagonal). The number of rules involved in the matrix generation, the number of elements combined in the matrix, and the logic used for combining together the rules determine the difficulty of the matrix. If the same rules, the same number of elements, the same logic but different elements (i.e., different shapes) are used for generating the matrices, the resulting matrices should have the same difficulty. The first aim of this study is to test this hypothesis by generating pairs of matrices that share the same rules, combined with the same directional logics, and the same number of elements, but displaying different shapes. If a matrix belonging to a pair is correctly (or incorrectly) responded, then the same should be observed on the other matrix belonging to the pair. The response that completes the matrix is presented along with (usually) 7 distractors. Since previous findings suggest that the correct response can be identified only by looking at the distractors (White & Zammarelli, 1981), the second aim of the study is to investigate the potential impact of the type (i.e., repetition, difference, wrong principle, incomplete correlate, Kunda et al. 2016) and the number of distractors on the difficulty of the matrices. To test this hypothesis, pairs of matrices with different number of distractors (i.e., 4 or 7) are generated. The 4 distractors are a subset of the 7 distractors. Hence, the proportion of each distractor type changes with the number of distractors.

• Stefanutti, L. (2019). On the assessment of procedural knowledge: From problem spaces to knowledge spaces. British Journal of Mathematical and Statistical Psychology, 72(2), 185-218. https://doi.org/10.1111/bmsp.12139
• Stefanutti, L., de Chiusole, D., & Brancaccio, A. (2021). Markov solution processes: Modeling human problem solving with procedural knowledge space theory. Journal of Mathematical Psychology, 103, 102552. https://doi.org/10.1016/j.jmp.2021.102552
• Brancaccio, A., de Chiusole, D., & Stefanutti, L. (2022) Algorithms for the adaptive assessment of procedural knowledge and skills. Behavior Research Methods

The basic local independence model (BLIM; Falmagne & Doignon, 1988) is a fundamental model in knowledge space theory. It describes the probability of observing a certain pattern of item responses as a function of the probability of the knowledge states (each of which being the set of all the items an individual is capable of solving) in the population and the careless error and lucky guess probabilities of the items. The estimates of these probabilities allow for uncovering the knowledge state of an individual from their response pattern. Different classes of individuals (e.g., young or old; with or without a specific clinical condition) exhibit similarities and specificities that should be taken into account in the assessment process. The project aims to develop an extension of the BLIM where the probabilities of the knowledge states and the careless error and lucky guess probabilities of the items are allowed to vary, independently of one another, between specific classes of individuals. Compared with the BLIM, the new model is expected to lead to more efficient and accurate assessments.