Exploring the Assistance Dilemma: The Case of Context Personalization

Exploring the Assistance Dilemma: The Case of Context Personalization Candace Ann Walkington (cwalkington@wisc.edu) Wisconsin Center for Educational Research, 1025 W. Johnson Street Madison, WI 53706 USA Keith Maull (Keith.Maull@colorado.edu) Institute of Cognitive Science, Department of Computer Science, 1777 Exposition Drive Boulder, CO 80301 USA mental effort. Cognitive load theory differentiates between extraneous cognitive load, stemming from activities not related to schema acquisition, intrinsic cognitive load, or inherent difficulty from interactivity of knowledge elements, and germane cognitive load, or effort related to schema acquisition (Sweller, Merrienboer, & Paas, 1998). Forms of assistance that reduce extraneous cognitive load should enhance learning by freeing up cognitive resources, if the schemas being learned are sufficiently challenging. However, also central to the assistance dilemma is the notion of desirable difficulties – research has shown that modifications that reduce performance during instruction (like decreasing feedback) can actually increase learning (Schmidt & Bjork, 1992), suggesting that mental effort can be germane to learning. The assistance dilemma considers when to give assistance and how much assistance to give in order to facilitate learning and learning efficiency, while also acknowledging that assistance can serve as a “crutch” or a “scaffold” (Koedinger et al., 2008). Abstract Context personalization, the practice of matching features of an instructional component to a learner‟s interests and experiences, has been framed in the literature as a means by which to improve learning by enhancing motivation. However, a related perspective could consider personalization a form of instructional assistance, with the potential to support the learning of new concepts. In this paper, the assistance dilemma, known to be “a fundamental unsolved problem in cognitive science” (Koedinger, Pavlik, McLaren, & Aleven, 2008, p. 2159), is investigated for context personalization. Two research studies explore whether personalization can be considered a form of assistance, and how this intervention mediates performance measures. Keywords: personalization; assistance dilemma; cognitive tutor; algebra Background The personalization hypothesis (learnlab.org/research/wiki) posits that matching up features of an instructional program to a learner‟s interests, experiences, or typical language usage will lead to more learning, compared to when instruction is not personalized. The benefits of personalization are often framed in terms of motivation or interest (e.g., Heilman, Collins, Eskenazi, Juffs, & Wilson, 2010); however personalization could also be considered a form of instructional assistance. This may be a useful perspective, given that research results for personalization are mixed, with some studies showing positive effects (Heilman et al., 2010) and others showing no effect (McLaren, Lim, Gagnon, Yaron, & Koedinger, 2006). Framing personalization as assistance makes explicit that benefits should only be expected in some cases, namely when the assistance is both needed and substantive. In this paper, the idea of personalization is explored from the perspective of an important issue in the science of instruction, the assistance dilemma. Context Personalization Here the primary topic of concern is one specific type of personalization - context personalization. In context personalization interventions, features of an instructional program are matched to individual learner‟s personal interests and experiences. For example, in mathematics, rather than being given a generic story problem on harvesting wheat from a field of grain, a learner might receive a variation of this problem based on their individual interests, perhaps a mathematical scenario about playing a video game or shopping at the mall. The idea that such personalization of mathematics problems may enhance learning is prevalent in the culture of schooling (Fives & Manning, 2005); however little research has empirically examined its impact. The assistance dilemma has sometimes been framed with respect to the “education wars,” or the struggle to strike a balance between giving more assistance (i.e. direct or traditional instruction) and less assistance (i.e. problem- solving and discovery learning) in learning environments. Context personalization is an especially fascinating instance of the assistance dilemma, because although it could be seen as a form of information-giving, it is widely supported by reform movements (e.g. National Council of Teachers of Mathematics, 2000). The Assistance Dilemma The assistance dilemma, or how to balance information- giving with information-withholding in learning environments, has been framed as “a fundamental unsolved problem in cognitive science” (Koedinger, Pavlik, McLaren, & Aleven, 2008, p. 2159). Assistance is considered not only to be direct hints or scaffolds, but any modification to the learning environment that enhances performance or reduces