How do Experts Adapt their Explanations to a Layperson’s Knowledge in Asynchronous Communication? An Experimental Study

Abstract

Despite a plethora of recommendations for personalization techniques, such approaches often lack empirical justification and their benefits to users remain obscure. The study described in this paper takes a step towards filling this gap by introducing an evidence-based approach for deriving adaptive interaction techniques. In a dialogue experiment with 36 dyads of computer experts and laypersons, we observed how experts tailored their written explanations to laypersons' communicational needs. To support adaptation, the experts in the experimental condition were provided with information about the layperson's knowledge level. In the control condition, the experts had no available information. During the composition of their answers, the experts in both conditions articulated their planning activities. Compared with the control condition, the experts in the experimental condition made a greater attempt to form a mental model about the layperson's knowledge. As a result, they varied the type and proportion of the information they provided depending on the layperson's individual knowledge level. Accordingly, such adaptive explanations helped laypersons reduce comprehension breakdowns and acquire new knowledge. These results provide evidence for theoretical assumptions regarding cognitive processes in text production and conversation. They empirically ground and advance techniques for adaptation of content in adaptive hypermedia systems. They are suggestive of ways in which explanations in recommender and decision support systems could be effectively adapted to the user's knowledge background and goals.

The effects of transparency on trust in and acceptance of a content-based art recommender

“…The ICC can range between 0 (indicating no interrater agreement) and 1 (indicating total interrater agreement). In this study, the intraclass correlation coefficient produced a value of 0.808, which can be considered to be very good Nückles et al (2006). Having obtained an acceptable interrater reliability, subsequent analysis was carried out on the data of the main coder only.…”

Conveying clinical reasoning based on visual observation via eye-movement modelling examples

“…One reason might be that the cues used are usually based on what area an instructional designer or domain expert thinks a novice learner at a particular moment should be looking at. Experts, however, may not necessarily know what the most important feature for novices is, since experts often have difficulties in correctly assessing the knowledge state of novices (Hinds 1999;Nückles et al 2006). …”

Layered evaluation of interactive adaptive systems: framework and formative methods

“…Nückles et al (2006) et al 1984). − Retrospective testing Using an interview or questionnaire, participants report their thoughts after the task has finished, possibly while watching a video of their actions (Nielsen 1994b).…”

“…For example, Nückles et al (2006) observed expert behaviour and how it was influenced by the availability of additional knowledge to them. Essentially, they were able to demonstrate which adaptation decisions were taken by experts given different user models.…”

“…This method was first fully described by Masthoff (2006), though it had been implicitly (partly) applied before (e.g., Masthoff 2004;Masthoff and Gatt 2006;de Rosis et al 2006;Nückles et al 2006). It integrates ideas from both contextual design and wizard-of-Oz.…”