Learner models hold and dynamically update the information about a user’s learning: current knowledge, competencies, misconceptions, goals, affective states, etc. There is an increasing trend towards opening the learner model to the user (learner, teacher or other stakeholders), often to support reflection and encourage greater learner responsibility for their learning; as well as helping teachers to better understand their students (Bull & Kay, 2010). A core requirement is that such visualisations must be understandable to the user. On the surface this may appear similar to the more recent work on learning analytics. However, open learner models (OLM) focus much more on the current state of learners, and with less reference to activities undertaken, scores gained, materials used, contributions made, etc. OLMs typically focus more on concepts or competencies, to guide learners towards consideration of conceptual issues rather than specific activities and performance.

Various OLM visualisation examples have been described in the literature for university students (see Bull & Kay 2010, for a more detailed overview). The most common visualisations used in courses include skill meters, concept maps and hierarchical tree structures. Recently, tree map overview-zoom-filter approaches to open learner modelling have also started to appear, as have tag/word clouds (Bull et al., in press; Mathews et al., 2012), and sunburst views (Conejo et al., 2011).

Girard and Johnson (2008) aimed to extend considerations about the presentation of OLMs to school level. To date, at that level most attempts have been very simple in presentation, a common example being smileys or proficiency indicated by colour (e.g. in Khan Academy, 2012). Other recent approaches include the many new learning analytics dashboards (e.g. Verbert et al., 2013), though these are not typically based on learner modelling. Research has considered whether some learner model visualisations may be more suitable than others. For example: a recent eye-tracking study of university students trying to understand four learner model presentations - kiviat chart, concept tag cloud, tree map and concept hierarchy - found the kiviat graph and concept hierarchy to be more efficient (for understanding the representation), than the tree map and tag cloud (Mathews et al., 2012).

Furthermore, the kiviat chart was considered to be the best format through which to gain a quick overview of knowledge; although other views were useful if further detail was required. Based on these results, Mathews et al (2012) conclude that the most useful visualisation is likely to depend on the context for which it is being used. Another OLM eye-tracking study found that, when concept map, pre-requisite map, tree structure of concepts, tree structure following lecture topics and sub-topics, and alphabetical index of the same underlying model were compared, visual attention in a view depended on whether the view was amongst the user’s preferred views, rather than on the structure of the view itself (Bull et al., 2007). Albert et al. (2010) state that university students thought simpler views were understandable and suitable for gaining an overview of their learning; while a more complex activity visualisation was less popular, perhaps because the complexity of the information was difficult to understand. Duan et al (2010) found preferences towards skill meters over more complex visualisations; however, users with more complex views of the learner model perceived it to be more accurate. Also, although a majority, the preference for the simpler skill meters was still only 53%. Ahmad and Bull (2008) also found that students perceived more detailed views of the learner model to be more accurate than skill meters. However, users still reported a higher level of trust in the more simple skill meter overviews. University students with experience with a range of different OLM systems have indicated their preference for having both overviews and detailed learner model information available for viewing (Bull, 2012).

Mr Collins was the first learner model designed for negotiation with the student (Bull & Pain, 1995). Its focus was on increasing model accuracy by student-system discussion of the model, while at the same time promoting learner reflection through discussion. The model contained separate belief measures: the system’s inferences about the student’s understanding, and the student’s own confidence in their knowledge, with identical interaction moves available to learner and system to try to resolve any discrepancies (e.g. challenge, offer evidence). Subsequent work extended the types of negotiation available, from the original menu-based discussion, to include dialogue games (Dimitrova, 2003) and chatbot (Kerly & Bull, 2008).

Close to the above definition of negotiated learner models is xOLM (Van Labeke et al., 2007). However, xOLM relied on the student to initiate discussion of the model. For example, students can challenge claims, warrants and backings, and receive justifications from the system. Learners can choose to agree, disagree, or move on (without resolution). Unlike full negotiation, the system allowed the learner's challenge to succeed where there is unresolved disagreement. In contrast, EI-OSM defers the overriding decision to the (human) teacher if interaction between a student and teacher cannot resolve discrepancies using the system’s evidence-based argument approach (Zapata-Rivera et al., 2007). There were mixed reactions from teachers as to whether they would consider assessment claims from students without the availability of relevant evidence, but they believed that these could form a useful starting point for formative dialogue.

While not a negotiated learner model, a Facebook group to allow university students to discuss their learner models with each other (Alotaibi & Bull, 2012), indicating a willingness to critically consider understanding in an open-ended way. This is crucial for methods of model negotiation between human partners where open discussion is encouraged. A similar approach in the sense that the model itself is not negotiated, is allowing children to provide their assessments of their knowledge to the system if they disagree with it, quantitatively and explained in text comments which can be seen by the teacher. Such input can become a focus for subsequent (human) teacher-child discussion (Zapata-Rivera & Greer, 2004). The NEXT-TELL CoNeTo tool (Vatrapu et al., 2012) offers a socio-cultural approach to negotiating the learner model, which allows learners and teachers to discuss the NEXT-TELL OLM (Bull et al., in press), while focussing on artefacts and evidence. In LEA’s BOX we aim to incorporate evidence-based discussion and learner model negotiation, at the level appropriate to age, STEM subject (with reference to competencies), and learning goals, as identified by teachers.