Ensuring that Machine Learning (ML) models make correct and meaningful inferences is necessary for the broader adoption of such models into high-stakes decision-making scenarios. Thus, ML model engineers increasingly use eXplainable AI (XAI) tools to investigate the capabilities and limitations of their ML models before deployment. However, explaining sequential ML models, which make a series of decisions at each timestep, remains challenging. We present Visual Interactive Model Explorer (VIME), an XAI toolbox that enables ML model engineers to explain decisions of sequential models in different “what-if” scenarios. Our evaluation with 14 ML experts, who investigated two existing sequential ML models using VIME and a baseline XAI toolbox to explore “what-if” scenarios, showed that VIME made it easier to identify and explain instances when the models made wrong decisions compared to the baseline. Our work informs the design of future interactive XAI mechanisms for evaluating sequential ML-based decision support systems.

Electrification is an important first step toward reducing the greenhouse emissions of passenger vehicles. However, how drivers drive, charge, and operate their electrified vehicles can have a large impact on their emissions, particularly for Plug-in Hybrid Electric vehicles (PHEVs) that combine all-electric driving with an internal combustion engine. In this paper, we investigate how and why drivers use their PHEVs and uncover design opportunities for interfaces that can support the efficient use of PHEVs. We used a mixed-method approach combining quantitative, qualitative, and concept elicitation methods with PHEV owners in the US. While past findings indicate that PHEV drivers are not motivated to charge regularly, our work contradicts this with evidence of (1) regular charging with home infrastructure, (2) high cost sensitivity, and (3) preference for driving in all-electric mode. Our results indicate that the most critical problem is inadequate user support for navigating poor charging infrastructure.

People consider recommendations from AI systems in diverse domains ranging from recognizing tumors in medical images to deciding which shoes look cute with an outfit. Implicit in the decision process is the perceived expertise of the AI system. In this paper, we investigate how people trust and rely on an AI assistant that performs with different levels of expertise relative to the person, ranging from completely overlapping expertise to perfectly complementary expertise. Through a series of controlled online lab studies where participants identified objects with the help of an AI assistant, we demonstrate that participants were able to perceive when the assistant was an expert or non-expert within the same task and calibrate their reliance on the AI to improve team performance. We also demonstrate that communicating expertise through the linguistic properties of the explanation text was effective, where embracing language increased reliance and distancing language reduced reliance on AI.

The demands of daily work offer few opportunities for workers to take stock of their own progress, big or small, which can lead to lower motivation, engagement, and higher risk of burnout. We present Highlight Matome, a personal online tool that encourages workers to quickly record and rank a single work highlight each day, helping them gain awareness of their own successes. We describe results from a field experiment investigating our tool's effectiveness for improving workers' engagement, perceptions, and affect. Thirty-three knowledge workers in Japan and the U.S. used Highlight Matome for six weeks. Our results show that using our tool for less than one minute each day significantly increased measures of work engagement, dedication, and positivity. A qualitative analysis of the highlights offers a window into participants' emotions and perceptions. We discuss implications for theories of inner work life and worker well-being.

Work breaks--both physical and digital--play an important role in productivity and workplace wellbeing. Yet, the growing availability of digital distractions from online content can turn breaks into prolonged "cyberloafing". In this paper, we present UpTime, a system that aims to support workers' transitions from breaks back to work--moments susceptible to digital distractions. Combining a browser extension and chatbot, users interact with UpTime through proactive and reactive chat prompts. By sensing transitions from inactivity, UpTime helps workers avoid distractions by automatically blocking distracting websites temporarily, while still giving them control to take necessary digital breaks. We report findings from a 3-week comparative field study with 15 workers. Our results show that automatic, temporary blocking at transition points can significantly reduce digital distractions and stress without sacrificing workers' sense of control. Our findings, however, also emphasize that overloading users' existing communication channels for chatbot interaction should be done thoughtfully.

Managing an audience's visual attention to presentation content is critical for effective communication in tele-conferences. This paper explores how audience and presenter coordinate visual and verbal information, and how consistent their gaze behavior is, to understand if their gaze behavior can be used for inferring and communicating attention in remote presentations. In a lab study, participants were asked first to view a short video presentation, and next, to rehearse and present to a remote viewer using the slides from the video presentation. We found that presenters coordinate their speech and gaze at visual regions of the slides in a timely manner (in 72% of all events analyzed), whereas audience only looked at what the presenter talked about in 53% of all events. Rehearsing aloud and presenting resulted in similar scanpaths. To further explore if it possible to infer if what a presenter is looking at is also talked about, we successfully trained models to detect an attention match between gaze and speech. These findings suggest that using the presenter's gaze has the potential to reliably communicate the presenter's focus on essential parts of the visual presentation material to help the audience better follow the presenter.

Work breaks can play an important role in the mental and physical well-being of workers and contribute positively to productivity. In this paper we explore the use of activity-, physiological-, and indoor-location sensing to promote mobility during work-breaks. While the popularity of devices and applications to promote physical activity is growing, prior research highlights important constraints when designing for the workplace. With these constraints in mind, we developed BreakSense, a mobile application that uses a Bluetooth beacon infrastructure, a smartphone and a smartwatch to encourage mobility during breaks with a game-like design. We discuss constraints imposed by design for work and the workplace, and highlight challenges associated with the use of noisy sensors and methods to overcome them. We then describe a short deployment of BreakSense within our lab that examined bound vs. unbound augmented breaks and how they affect users' sense of completion and readiness to work.

Video telehealth is growing to allow more clinicians to see patients from afar. As a result, clinicians, typically trained for in-person visits, must learn to communicate both health information and non-verbal affective signals to patients through a digital medium. We introduce a system called ReflectLive that senses and provides real-time feedback about non-verbal communication behaviors to clinicians so they can improve their communication behaviors. A user evaluation with 10 clinicians showed that the real-time feedback helped clinicians maintain better eye contact with patients and was not overly distracting. Clinicians reported being more aware of their non-verbal communication behaviors and reacted positively to summaries of their conversational metrics, motivating them to want to improve. Using ReflectLive as a probe, we also discuss the benefits and concerns around automatically quantifying the "soft skills" and complexities of clinician-patient communication, the controllability of behaviors, and the design considerations for how to present real-time and summative feedback to clinicians.

Medication taking is a self-regulatory process that requires individuals to self-monitor their medication taking behaviors, but this can be difficult because medication taking is such a mundane, unremarkable behavior. Ubiquitous sensing systems have the potential to sense everyday behaviors and provide the objective feedback necessary for self-regulation of medication taking. We describe an unobtrusive sensing system consisting of a sensor-augmented pillbox and an ambient display that provides near real-time visual feedback about how well medications are being taken. In contrast to other systems that focus on reminding before medication taking, our approach uses feedback after medication taking to allow the individual to develop their own routines through self-regulation. We evaluated this system in the homes of older adults in a 10-month deployment. Feedback helped improve the consistency of medication-taking behaviors as well as increased ratings of self-efficacy. However, the improved performance did not persist after the feedback display was removed, because individuals had integrated the feedback display into their routines to support their self-awareness, identify mistakes, guide the timing of medication taking, and provide a sense of security that they are taking their medications well. Finally, we reflect on design considerations for feedback systems to support the process of self-regulation of everyday behaviors.

Older adults often struggle with maintaining self-aware of their ability to carry out everyday activities important for independence. Unobtrusive sensors embedded in the home can monitor how older adults interact with objects around the home and can provide objective accounts of behaviors to support self-awareness. In this paper, we describe the design and four month deployment of a prototype sensing system that tracks medication taking and phone use in the homes of two older adults. We describe two case studies on 1) how they engaged with the data by looking for and explaining their own anomalous behaviors and 2) how they used the sensor data to reflect on their actions and their own self-awareness of their abilities to remain independent. Finally, we propose recommendations for the design of home sensing systems that support awareness of functional abilities for older adults using reflection.

Lifelogging technologies have the potential to provide memory cues for people who struggle with episodic memory impairment (EMI). These memory cues enable the recollection of significant experiences, which is important for people with EMI to regain a sense of normalcy in their lives. However, lifelogging technologies often collect an overwhelmingly large amount of data to review. The best memory cues need to be extracted and presented in a way that best supports episodic recollection. We describe the design of a new lifelogging system that captures photos, ambient audio, and location information and leverages both automated content/context analysis and the expertise of family caregivers to facilitate the extraction and annotation of a salient summary consisting of good cues from the lifelog. The system presents the selected cues for review in a way that maximizes the opportunities for the person with EMI to think deeply about these cues to trigger memory recollection on his own without burdening the caregiver. We compare our system with another review system that requires the caregiver to repeatedly guide the review process. Our self-guided system resulted in better memory retention and imposed a smaller burden on the caregiver whereas the caregiver-guided approach provided more opportunities for caregiver interaction.

Alzheimer's disease impairs episodic memory and subtly and progressively robs people of their ability to remember their recent experiences. In this paper, we describe two studies that lead to a better understanding of how caregivers use cues to support episodic memory impairment and what types of cues are best for supporting recollection. We also show how good memory cues differ between people with and without episodic memory impairment. We discuss how this improved understanding impacts the design of lifelogging technologies for automatically capturing and extracting the best memory cues to assist overburdened caregivers and people with episodic memory impairment in supporting recollection of episodic memory.