Further results on gender and cognitive differences in help effectiveness Ivon Arroyo 1 , Tom Murray 1,2 , Beverly P. Woolf 1 , Carole R. Beal 1 1 University of Massachusetts, Amherst - 2 Hampshire College, Amherst Abstract. We explored the effectiveness of help for 350 students of different genders and cognitive developments, in an arithmetic intelligent tutoring system. We conclude that girls were more sensitive to the amounts of help fitting their needs than to the level of abstraction. On the other hand, boys were affected by the abstraction level, and ignored help more. Introduction Past research suggests that what constitutes good teaching will not be the same for students of different genders [Carr, 97; Royer, 99] and cognitive developments [Suydam&Higgins,77; Fuson&Briars,90]. Cognitive development research suggests that concrete-operational students benefit from concrete teaching materials more than numeric-procedures to learn arithmetic. Meanwhile, the two genders seem to choose different approaches to problem solving, girls using more concrete and overt strategies, boys using covert approaches and retrieval from memory. These concepts can be extended to the kinds of help that the two genders need from educational software. However, questions arise when looking at the interactions of cognitive development and gender research. Research suggests that providing concrete help would benefit concrete students. Still, it is not clear how ineffective would be to provide low cognitive development girls abstract help with high amounts of structure and interactivity, or how beneficial concrete representations in the help would be for low cognitive development boys when provided in a structured interactive fashion. This paper extends previous results [Arroyo, 00] to provide a thorough analysis of the interaction between these two student characteristics, with a controlled evaluation involving 350 students. We explore how differently the genders and students of different cognitive developments reacted to the help in our mathematics intelligent tutor, AnimalWatch. Four different versions of AnimalWatch were evaluated. This paper analyzes help impact along different measures. It then analyzes patterns across them to decide what would be the best kind of help to provide girls, boys, concrete operational and formal operational students. 1. AnimalWatch AnimalWatch is an Intelligent Tutoring System for whole number arithmetic and fractions, for 9-11 year olds [Beal, 02]. Four different versions of AnimalWatch have been developed, which vary in the kind of immediate help provided when students make mistakes. In a reduced help version, students get short help messages providing little feedback. In a concrete manipulatives version, students see interactive hints involving concrete manipulatives (base-10 blocks, bars for fractions). In a formal version, the help emphasizes numeric procedures, where interactive numeric hints are provided. In a concrete+formal version, students get both kinds of hints, one after the other. All versions of AnimalWatch eventually “bottom-out” the help by providing the correct answer. 2. Experiment design Cognitive development was determined with a computer-based cognitive development pre- test, which tests for concrete and formal-operational thought [Arroyo, 01]. Students were