Stanford CS547 HCI Seminar | Spring 2026 | Toward Ontological Multiplicity in AI and Computing

AI development often relies on narrow, hidden assumptions about the world, which restricts what systems can do and how users understand their lives. By shifting toward "ontological multiplicity," designers can create software that supports fluid, evolving categories and relationships instead of forcing data into rigid, binary structures.

• Hidden worldviews — Software development encodes unstated beliefs into code, which dictates how users interpret their own lives and environment [01:52].

• Glitch analysis — Monitoring skin conductance showed that "non-responsive" users are usually excluded; however, focusing on personal events revealed "purple zones," an ambiguous state where boundaries blur [11:21].

• User-led labeling — Testing allowed participants to define their own data categories, which resulted in meaningful, personal labels—like "chill walk"—rather than standard metrics [25:01].

• Design shifts — Technical practice should evolve by moving away from binary classification toward:

  • Relational focus — Prioritizing connections between entities rather than forcing items into isolated boxes [17:48].
  • Cultivating change — Influencing environments rather than demanding direct control over automated outcomes [18:03].
  • Algorithmic care — Valuing the "in-between" spaces where boundaries are negotiated, rather than just forcing clear-cut visibility [18:40].

• Analytical framework — Applying four lenses—multiplicity, groundedness, liveliness, and enactments—allows teams to identify and challenge the hidden limits embedded in systems [30:44].

• How does incorporating "purple zone" or "glitchy attunement" concepts into machine learning development affect the measured reliability of resulting models?

• What are the practical challenges when designing systems that prioritize relational meaning-making over standardized user metrics?