UMBC Interactive Systems Research Center

Permanent URI for this collectionhttp://hdl.handle.net/11603/11278

The Interactive Systems Research Center (ISRC) acts as a bridge for faculty across the UMBC campus with expertise in designing, building, or studying uses of interactive computing systems. It facilitates the sharing of resources and experience in solving computing problems from a user-centered perspective grounded in user needs and not in simply applying previously designed solutions to new domains.

The ISRC also acts as a bridge to industry, community, and organizations in order to foster long-standing relationships in addressing user needs through interactive computing solutions. Through the center, external collaborators can seamlessly work with a mix of expertise to innovate interactive computing solutions of societal importance.

Browse

Recent Submissions

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    GPT's Devastated and LLaMA's Content: Emotion Representation Alignment in LLMs for Keyword-based Generation
    (2025-03-14) Choudhury, Shadab Hafiz; Kumar, Asha; Martin, Lara J.
    In controlled text generation using large language models (LLMs), gaps arise between the language model's interpretation and human expectations. We look at the problem of controlling emotions in keyword-based sentence generation for both GPT-4 and LLaMA-3. We selected four emotion representations: Words, Valence-Arousal-Dominance (VAD) dimensions expressed in both Lexical and Numeric forms, and Emojis. Our human evaluation looked at the Human-LLM alignment for each representation, as well as the accuracy and realism of the generated sentences. While representations like VAD break emotions into easy-to-compute components, our findings show that people agree more with how LLMs generate when conditioned on English words (e.g., "angry") rather than VAD scales. This difference is especially visible when comparing Numeric VAD to words. However, we found that converting the originally-numeric VAD scales to Lexical scales (e.g., +4.0 becomes "High") dramatically improved agreement. Furthermore, the perception of how much a generated sentence conveys an emotion is highly dependent on the LLM, representation type, and which emotion it is.
  • Thumbnail Image
    Item
    An Ecosystem of Support: A U.S. State Government-Supported DIY-AT Program for Residents with Disabilities
    (ACM, 2024-10-27) Higgins, Erin; Sakowicz, Marie E.; Hamidi, Foad
    While Do-It-Yourself (DIY) approaches to producing customized assistive technology (AT) have been shown to support end-user agency and further technology democratization, research has shown that the utilization of digital fabrication tools requires a high level of technical expertise as well as financial investment. Facilitation of collaborations between end users and makers is a possible solution to these issues, however, previous efforts have uncovered issues around shared language, lack of consistent communication, and liability concerns. A promising direction for addressing these issues is to conceive of new types of multi-organizational collaborations that draw on complementary strengths. We explored these possibilities through an Action Research study in which we collaborated with the Maryland department of disability to launch a state level DIY-AT program. Through developing and supporting this program, we studied motivations for participation, relationships to creating and customizing AT, how individuals participated in and grew the program, and how the program allowed for individuals to reflect on their disabilities and AT use. Our findings generated an ecosystem model describing the interdependent relationships between and roles held by each stakeholder in the state DIY-AT program as well as a description of how this ecosystem encouraged expanding and transcending the understandings and definitions of AT and disability. We offer lessons learned for the design of future government-supported DIY-AT programs and reflections on the role of HCI researchers within these ecosystems.
  • Thumbnail Image
    Item
    Decoding the Privacy Policies of Assistive Technologies
    (ACM, 2024-10-22) Crawford, Kirk; Khoo, Yi Xuan; Kumar, Asha; Mentis, Helena; Hamidi, Foad
    As assistive technologies (ATs) have evolved, they have become increasingly connected. However, these increasing connections pose significant privacy challenges, especially when user privacy is described using complex privacy policies. Our study decodes the privacy policies of 18 ATs to understand how data collection and processing are communicated with users. We find that (1) AT privacy policies are structured to offer legal protections to their companies and not always to protect user privacy, (2) AT privacy policies are absent protections for individuals with disabilities, (3) AT policies are inconsistent when describing data storage, handling, and security methods, (4) AT policies often do not differentiate between essential and non-essential data collection, and (5) there is often a lack of transparency in AT policies around third-party data sharing. These findings reveal that AT privacy policies overlook and underestimate a user?s acceptable privacy risks. We conclude our study by discussing AT design implications.
  • Thumbnail Image
    Item