Lev Tankelevitch’s research while affiliated with Cancer Research UK Cambridge Institute and other places

... Additionally, more and better synthesized design-oriented knowledge that captures how to design for human agency, oversight, and AI literacy is pending. To date, research is revisiting the value of these design principles, e.g., in the context of AI (Long and Magerko, 2020) or generative AI (Simkute et al., 2024;Weisz et al., 2024). However, these design principles need clearer examples of how to operationalize them concretely within applications. ...

Reference:

Establishing and evaluating trustworthy AI: overview and research challenges

... As Generative AI (GenAI) tools achieve widespread penetration in knowledge workflows, an increasingly common observation made by user studies is that prompting, i.e., the skillful and effective use of natural language for steering and eliciting GenAI behavior, is challenging for many users in many contexts [11,34,60, 66]. The challenges range from the increased explicit metacognitive demands of prompting, to the burden of explaining detailed requirements and context in textual format, to the opaque and unpredictable mapping of naturalistic language queries to system behavior. ...

Reference:

Dynamic Prompt Middleware: Contextual Prompt Refinement Controls for Comprehension Tasks

... The clear definition of objectives, agendas, preparatory materials, and task-related resources is an essential aspect of efficient meeting preparation and facilitation, notably in virtual settings [19,60]. Yet, due to time constraints, both meeting planners and participants may cut corners in laying out pre-meeting preparations [92]. During the meeting, while explicit items on the agenda direct the discourse, the meeting actually often transitions through implicit phases. ...

Reference:

The CoExplorer Technology Probe: A Generative AI-Powered Adaptive Interface to Support Intentionality in Planning and Running Video Meetings

... Penultimately, Tiffany did not engage with Alex-TA outside of the first interview, their non-engagement rationale resonates with other work we have done wherein students who were learning something they perceived as complex did not have the cognitive bandwidth to also use Alex-TA-a complex process in its own right requiring a level of task and tool meta-awareness they might develop over time but not have initially. [33,34] The integration of AI tools into the curriculum to complement and transform traditional teaching and learning methods appears both advisable and inevitable. In doing so, it is crucial to promote responsible AI use and ensure that the technology supports rather than replaces critical thinking, student growth, skill development, and intellectual advancement. ...

Reference:

PERSPECTIVES AND EXPERIENCES OF USING GENERATIVE AI BY UNDERGRADUATE SPEECH LANGUAGE PATHOLOGY STUDENTS FOR COURSE NAVIGATION AND CAREGIVER INFORMATION SESSION PREPARATION

... Other work showed that automatic detection of different types of conversations [116], emotions [117], bodily cues [16], and environmental factors [20] in meetings can predict aspects of meeting effectiveness. For reducing post-meeting effort, significant research has gone into automatic generation of meeting summaries, highlights, and action items, [4,54,74,75], as well as metrics [21,43] and post-meeting dashboards that attempt to provide feedback on meeting effectiveness [86,87] Both the prior research and commercial tools show promise, but they have shortcomings if we consider the overall problems of ineffective meetings and meeting fatigue. First, they concentrate on features to improve specific pain points, usually because there is a clear technical path to a solution (e.g. ...

Reference:

The CoExplorer Technology Probe: A Generative AI-Powered Adaptive Interface to Support Intentionality in Planning and Running Video Meetings

... Head-tracked speakers or headphones allow us to simulate locations of sound sources around the user through spatial audio (e.g., placing a sound source in mid-air [34]). Using spatial audio for speech has been shown to aid speaker identification, improve intelligibility, reduce cognitive load, and increase social presence [2,37,43,50] and spatialized environmental sounds can lead to increased spatial awareness of objects in the space [58]. ...

Reference:

There Is More to Avatars Than Visuals: Investigating Combinations of Visual and Auditory User Representations for Remote Collaboration in Augmented Reality

... Such an observation is then used to improve the satisfaction of the users through the attraction paradigm, i.e., the tendency to like robots according to whether their traits are perceived to be similar or different [6]. Similarly, findings related to the proxemics of virtual environments can play a significant role in the the improvement of human-agent interactions [1] or the development of voice spatialisation in technology-mediated communications such as video conferencing [19] and hybrid learning [50]. ...

Reference:

Is Distance a Modality? Multi-Task Learning for Speech-Based Joint Prediction of Attributed Traits and Perceived Distances in 3D Audio Immersive Environments