The main research question revolves around how to design mathematics software that is effective at improving affective outcomes as well as achievement for all students, by responding to students' affective states within mathematics intelligent tutoring systems.
As part of this grant, we will first validate current existing models of emotional states within Intelligent Tutoring Systems for math. Then, several empirical experiments will unveil whether concrete cognitive, meta-cognitive and/or affective interventions can change students’ negative feelings (activating and deactivating), compared to control groups that receive partial or no emotion-sensitive support.
To address these questions, we will conduct nine experiments in Massachusetts and Arizona public schools, 7th through 10th grades. Schools are representative of different SES, equal male-female ratio. We will oversample low achieving students by working with remedial math classes whenever possible, to increase the chances of getting students with frequent negative emotions. For all experiments, outcome measures and quantitative and qualitative measures will be similar. Each experiment will engage approximately 100 students, from remedial math classes in each of the participating schools.
The project uses a cross-sectional and comparative research design and will generate evidence that is descriptive [design research and randomized/controlled] and causal [experimental]. Original data will be collected on 7th to 10th grade students in math classes in public schools, oversampling low achieving student populations, using assessments of learning and survey research [self-completion questionnaires and online].
Nine variations of the math tutoring system, addressing affect, context and meta-cognition, will be compared to the same system without those additions. Instruments or measures being used include the Achievement Emotions Questionnaire for mathematics, AEQ-M, by Pekrun, Goetz, Frenzel (2005) University of Munich. The project will use Analysis of Variance for the affective repairs experiments. For the initial validation phase, a variety of techniques are applied, including linear regression, classifiers and pattern detection.
Findings will be posted as they become available.
Freely available tutoring software that repairs negative emotional states when learning mathematics, called Wayang Outpost.