Millis, K.K.(PI), Britt, A., Magliano, J., Wiemer-Hasting, K., Graesser, A.C. (co-PI), Halpern, D. Acquiring research investigative and evaluative skills (ARIES) for scientific inquiry, Institute of Education Sciences, 2007-2011, $640,000 (Memphis allocation).

We envision ARIES: a state of the art intelligent tutor that teaches vital concepts of scientific inquiry by having the user hold conversations with two animated pedagogical agents as he or she solves a number of engaging problems in the social and physical sciences. Users will read an online text describing and explaining key concepts in scientific inquiry. To promote deep learning, users will teach an animated Other-Agent as the Guide-Agent looks on and makes suggestions. Later, the user will apply the concepts to problems that require the critical evaluation of realistically presented studies and causal claims. The Guide-Agent will tutor the user in natural language by using AutoTutor, an effective computerized tutor that mimics the dialog moves between human tutors and students. Users will be awarded points based on their answers and progress, giving the learning sessions the feel of a serious game. Every aspect of ARIES has been designed to promote the deep learning of scientific inquiry by implementing well known principles of learning. These include reciprocal teaching, self-explanation, spaced learning, practice at retrieval (testing effects), authentic learning, formative feedback, active responding, reflection, dialog interactivity and variable encoding.

Research will be conducted in a rural university in Illinois. The majority of participants will be college students with a diverse race/ethnicity and SES backgrounds. This proposal includes five randomized controlled experiments and one quasi-experiment, each testing a particular aspect of ARIES. Experiment one will examine the extent that having problems from different disciplines will produce transfer. Another will compare the relative effectiveness of role-playing on different types of problems. The third will examine whether full dialog exchanges for the problems are necessary for deep learning. The fourth will test the effectiveness of reciprocal teaching using animated agents. The fifth will test whether the effectiveness of reciprocal teaching depends on the knowledge level of the teachable agent. Finally, the success of ARIES on learning will be tested in university courses in which one class section will assign ARIES as homework while the other serves as a no-treatment control. The experiments will use many of the same measures: multiple-choice, evaluate studies, and design studies. Pre-and post-versions of each will be prepared. The rationale for using these various types of tests is that each assesses a different type of knowledge or retrieval process. Multiple-choice questions tap knowledge via recognition memory; evaluate problems measure the degree that the student can recall and use the knowledge in an authentic way; design problems measure the active procedural knowledge relevant to scientific inquiry. The data will be analyzed by ANOVA, ANCOVA, and hierarchical linear regression.

ARIES is a practical solution to a serious problem in science education. It appeals to administrators because of its low cost. It appeals to science teachers because it gives them the flexibility to use it in the classroom or as assigned homework. These educators will be pleased to know that ARIES uses effective tutoring technology. With ARIES on the Internet, high school and college students and other interested individuals will have access to learning scientific inquiry in an engaging way. It has the potential to increase students¡¯ interest in pursuing scientific careers. ARIES is the key to a flexible low-cost solution for learning scientific inquiry.