This paper describes methods for comparative evaluation of the interpretability of models of high dimensional time series data inferred by unsupervised machine learning algorithms. The time series data used in this investigation were logs from an immersive simulation like those commonly used in education and healthcare training. The structures learnt by the models provide representations of participants' activities in the simulation which are intended to be meaningful to people's interpretation. To choose the model that induces the best representation, we designed two interpretability tests, each of which evaluates the extent to which a model’s output aligns with people’s expectations or intuitions of what has occurred in the simulation. We compared the performance of the models on these interpretability tests to their performance on statistical information criteria. We show that the models that optimize interpretability quality differ from those that optimize (statistical) information theoretic criteria. Furthermore, we found that a model using a fully Bayesian approach performed well on both the statistical and human-interpretability measures. The Bayesian approach is a good candidate for fully automated model selection, i.e., when direct empirical investigations of interpretability are costly or infeasible.

Immersive open-ended museum exhibits promote ludic engagement and can be a powerful draw for visitors, but these qualities may also make learning more challenging. We describe our efforts to help visitors engage more deeply with an interactive exhibit's content by giving them access to visualizations of data skimmed from their use of the exhibit. We report on the motivations and challenges in designing this reflective tool, which positions visitors as a "human in the loop" to understand and manage their engagement with the exhibit. We used an iterative design process and qualitative methods to explore how and if visitors could (1) access and (2) comprehend the data visualizations, (3) reflect on their prior engagement with the exhibit, (4)plan their future engagement with the exhibit, and (5) act on their plans. We further discuss the essential design challenges and the opportunities made possible for visitors through data-driven reflection tools.

The wide availability of body-sensing technologies (such as Nintendo Wii and Microsoft Kinect) has the potential to bring full-body interaction to the masses, but the design of hand gestures and body movements that can be easily discovered by the users of such systems is still a challenge. In this paper, we revise and evaluate Framed Guessability, a design methodology for crafting discoverable hand gestures and body movements that focuses participants' suggestions within a "frame," i.e. a scenario. We elicited gestures and body movements via the Guessability and the Framed Guessability methods, consulting 89 participants in-lab. We then conducted an in-situ quasi-experimental study with 138 museum visitors to compare the discoverability of gestures and body movements elicited with these two methods. We found that the Framed Guessability movements were more discoverable than those generated via traditional Guessability, even though in the museum there was no reference to the frame.

Zoos and museums often rely on interpretive staff, called explainers, to facilitate visitors' learning through conversations and demonstrations. Many explainers begin their careers as teens, and would benefit from ongoing Professional Development (PD). As institutions begin to use mobile devices to enhance explainers' interpretation, new opportunities arise to support explainers' individual and collaborative professional development. This paper presents the results of structured participatory design sessions to engage explainers in examining and proposing features for a Facilitation, Reflection, and Augmented Interpretation Mobile System (FRAIMS). The goal for FRAIMS is to support everyday interpretive tasks while also gathering information on how explainers perform that interpretation (both passively, via logging and recording, and actively, via self-reports and ratings) to support them in their PD. Reflecting on one's own performance and the performance of others is a powerful PD strategy, but can be emotionally fraught. Via participatory design sessions with expert, in-development, and novice explainers at different informal learning institutions, we found that explainers' preferences for socially sharing performance information gathered via mobile devices varied with their experience. We detail emerging themes captured from the sessions and make suggestions for how these findings might apply more broadly to computer-supported professional development systems.

Interactive technologies like multi-touch tables enable museum exhibit designers to support visitors' learning with a wide range of resources (from multimedia to dynamic feedback to the presence of other visitors). Designers must decide, though, how best to align the affordances of these resources with the learning activities they are trying to support [23]. This work examines a multi-touch table exhibit designed to support an activity, tinkering, which has been identified as a form of interaction that may offer special benefits to novices learning about engineering [e.g., 4]. When a learner is tinkering, he or she is engaged in a process of iterative adjustment to a constructed artifact, making use of "just in time" resources and feedback to guide the next steps in their exploration of the problem space. The exhibit studied in this work provided several resources for supporting tinkering, and this paper presents a detailed case showing how these different resources (some technical, some social, and some sociotechnical) were or were not used by learners. A key design goal we identified was the need to transform the tacit engineering practices of visitors into visible engineering performances, such that those performances could serve as "cultural tools" [35] for mediating the learning of other visitors.

Informal science institutions (ISIs) are beginning to adopt mobile technology to support interpreters (docents), but not much is known about how to design these supports. One approach to designing technology for new scenarios is participatory design (PD), where end-users are involved as experts in the task domain who can help envision the application of technology. However, in our context end-users are often youth interpreters who are emerging professionals. This poses a challenge because traditional PD methods trust that the users can represent the task domain. Novice professionals may not yet fully understand the task domain, but eliciting their needs and visions is still important for producing a tool they will find useful. A design approach is needed that captures the requirements for supporting expert task execution as an underlying structure for the tool, while nonetheless eliciting and respecting the special needs of novices. We developed and applied two different framing strategies (one technological, one sociotechnological) to traditional PD methods to help youth non-expert interpreters generate task-relevant design ideas. We report results from using these strategies in an exploratory fashion and discuss opportunities for future research on PD methods that can address the needs of youth as emerging professionals.

Temporal reasoning (i.e., reasoning about relationships across time) is complex and difficult, particularly when engaged through complex media such as online Geographic Information System (GIS) applications. Partnering with Social Explorer (SE), a Web-based GIS application that allows users to create interactive visualizations of large sociological datasets, we engaged in frequent sequential pattern mining of a database of users' interactions with SE. The resulting frequent sequences provide initial descriptions of how SE affords opportunities to engage in temporal reasoning.

The "Uncanny Valley" theory explains the counter-intuitive phenomenon where people may get suddenly uncomfortable with an artificial entity when it becomes very similar to humans. We propose the existence of an "uncanny valley" for embodied interaction, when a user's body motions in the physical space (the locus of interaction) are incompletely mapped into effects in the virtual space (the focus of interaction). It is generally assumed that this mapping should be as veridical as possible to promote seamless embodied interaction. Many design factors (e.g., synchronicity, sensitivity, shared realism) contribute to veridical locus-focus mapping. We intentionally varied the level of veridicality of these different factors, affecting how the user's movements were mapped to virtual effects. Our results indicate that there is a dip (valley) in user preferences when the design contains mixed degrees of veridicality. Thus, when one veridical dimension is limited, designers should likewise reduce the veridicality of other dimensions.

We describe the development and user testing of an iPad prototype for a new museumbased, interactive tabletop computer game called MakeScape Lite. The game lets learnerswho have no prior formal experience in circuitrybuild virtual circuits in a fantasy challenge scenario to (1) learn some basic circuitry concepts and (2) engage in basic engineering practices, such as problem identification, design, and iteration. In phase 1 of testing, we explored how children socially interacted to develop design goals when the game lacked explicit guidance about what to do. In phase 2, which we discuss here, we iterated the prototype with partially worked examples to offer more explicit guidance to children. We then studied video and game log data of students playtesting the game.

As data rather than physical artifacts become more commonly the product of modern scientific endeavor, we must attend to humandata interactions as people reason about and with representations of data increasingly being presented in museum settings. Complex data sets can be impenetrable for novices, so the exhibit presented here was designed to give visitors control over a personalized "slice" of the data set as an entry point for exploration. Personalized control and collaboration can often be at odds in exhibits, however. This paper presents a study of two alternate approaches to designing an embodied interaction control for the exhibit that serves both needs. The results demonstrate that interaction design can affect children's perspective taking as they interact with a Census data map museum display, and that the perspective taken by individuals is correlated with their operation of the interactive exhibit and the kinds of reasoning they employ while investigating data.

The propagation of Inquiry Based Learning has lead to many more elementary students interacting with authentic scientific tools and practices. However, the more problematic realities of scientific data collection, such as noise and large data sets, are often deliberately hidden from students. Students will need to confront these realities and be able to make skillful data scoping decisions in order to make sense of ever more prevalent large datasets. We dub software designed to support these activities Information-Building Applications (IBAs). This paper presents the design considerations that went into building an exemplar IBA, PhotoMAT (Photo Management and Analysis Tool), a brief user study to show how the solutions enacted by following these principles are taken up by actual students, and a discussion of how the design considerations identified by our work might be applied to another IBA.

Despite the wide availability of body-sensing technologies, the design of control gestures that feel natural and that can be intuitively "guessed" by the users is still an embodied interaction challenge. This is especially true for systems that require a set of complementary control gestures. Part of the problem lies in the separation between the locus of the interaction (the body) and the focus of the interaction (the screen). We extend Johnson's theory of Embodied Schemata with Embodied Allegories, in order to create a unifying context that spans across the locus and focus of interaction. We present results that demonstrate how this approach increases the chance that users select the same gesture or movement for producing an effect within the virtual context, and that the resultant gesture set is deemed more intuitive by users. We also present the accompanying methodology, "Framed Guessability," which can increase users' agreement when conducting Guessability Studies.

This paper describes an empirical study conducted to explore the expressive design space for what we dub "effortful interaction." Effortful interaction is a form of human-computer interaction where different degrees of physical effort are intentionally incorporated into the interaction design in order to communicate information. In this sense effortful interaction is similar to haptics, where devices provide physical force feedback to users, except the perceived feedback occurs entirely within users' proprioception: i.e., users' perception of their own bodies. Our larger design goal is to use effortful interaction to communicate quantitative information to children who aren't yet proficient with graphs. First, though, we had to establish if we could reliably induce sensations of effort. This study examined the sensitivity of children to three different design parameters (the duration of interaction, the intensity required by the interaction, and the physical modality of the interaction), and compares their responses to those of adults.

Museums are increasingly embracing technologies that provide highly-individualized and highly-interactive experiences to visitors. With embodied interaction experiences, increased localization accuracy supports greater nuance in interaction design, but there is usually a tradeoff between fast, accurate tracking and the ability to preserve the identity of users. Customization of experience relies on the ability to detect the identity of visitors, however. We present a method that combines fine-grained indoor tracking with robust preservation of the unique identities of multiple users. Our model merges input from an RFID reader with input from a commercial camera-based tracking system. We developed a probabilistic Bayesian model to infer at run-time the correct identification of the subjects in the camera's field of view. This method, tested in a lab and at a local museum, requires minimal modification to the exhibition space, while addressing several identity-preservation problems for which many indoor tracking systems do not have robust solutions.

This paper describes a frequently-overlooked aspect of embodied interaction design: physical effort. Although exertion is the direct goal of many embodied activities (e.g., exergames), and is used indirectly to discourage certain user interactions (as with affordances), exertion has not been used to support direct expressive interaction with an embodied system. Situating exertion in both psychological and physiological literature, this paper suggests guidelines for employing exertion as more than just an incidental component of proprioception in embodied interaction designs. Specifically, the linkages between exertion, affect, and recall are reviewed and analyzed for their potential to support embodied learning activities, and literature concerning human perceptions of effort is reviewed to help designers understand how to incorporate effort more directly and intentionally in embodied interaction designs. Also presented is an illustration of how these guidelines affected the design of an educational embodied interaction experience for an informal learning setting.

New mobile device features and the growing proportion of visitors carrying mobiles allow the range of museum exhibit design possibilities to be expanded. In particular, we see opportunities for using mobiles to help exhibits scale up to support variable-sized groups of visitors, and to support collaborative visitor-visitor interactions. Because exhibit use is generally one-time-only, any interfaces created for these purposes must be easily learnable, or visitors may not use the exhibit at all. To guide the design of learnable mobile interfaces, we chose to employ the Consistency design principle. Consistency was originally applied to desktop UIs, so we extended the definition to cover three new categories of consistency relevant to ubiquitous computing: Within-Device Consistency, Across-Device Consistency and Within-Context Consistency. We experimentally contrasted designs created from these categories. The results show small differences in learnability, but illustrate that even for one-off situations learnability may not be as important as usability.

Informal learning environments (e.g. children's science museums) provide challenges for educational software design that are distinct from those found in traditional classrooms. Predominant among them is the need to support groups of users in heavily social learning experiences. In prior work, we have found that providing mobile, wirelessly-linked devices to individual participants allows multiple simultaneous users to learn together with a shared simulation [17, 18, 15, 27]. To extend this design strategy to an informal environment like a museum, however, other environmental differences (e.g. expected length of participation) must be taken into account. I propose the use of dynamic and individually-targeted software-based scaffolding to help groups of visitors to more productively experiment with a shared simulation. I utilize a design experiment approach to focus on the design and evaluation of these distributed scaffolds that enable informal learners to cooperatively participate in a joint simulation.