CHAPTER 12 – Methods for Assessing Inquiry : Machine - learned and Theoretical

Inquiry skills are critical to virtually any kind of problem solving and particularly to the practice of science. However, these skills cannot be easily assessed using traditional static assessment items, in which it is difficult to follow (and capture) the actual inquiry process. Given a sufficiently interactive task, the challenge becomes how to identify and score the inquiry skills applied to the task. This chapter examines three different methods for assessing inquiry skills using the process data generated by computerized interactive tasks. The first two methods identify inquiry strategy-use based on specific features in the data. The first uses theoretically defined features, whereas the second leverages machine learning to identify and combine relevant features. The third method uses generative process models to compare student actions to probabilistic agents implementing targeted inquiry strategies. The advantages and drawbacks of each method are discussed along with assessment contexts that favor particular approaches. Implications for detection and scoring of inquiry within the Generalized Intelligent Framework for Tutoring (GIFT) are addressed.