What really separates designing for people's needs from designing for their deeper values, especially when AI is involved?
What makes value-sensitive design distinct from user-centered design approaches?
This explores what sets value-sensitive design (which builds human values like fairness, privacy, or autonomy into a system from the start) apart from user-centered design (which builds around what users need and how they behave), and what that difference means for AI.
This explores what separates designing around people's values from designing around users' needs, especially for AI systems. The corpus has little on value-sensitive design head-on, so this is a partial answer, but the material it does have points to a useful distinction. The most direct doorway is a mapping of eight participatory traditions that shape AI design What participatory traditions shape AI design practice?. It places user-centered design and value-sensitive design side by side with participatory design and co-design, and it separates them by their goals and by how much power each gives the people affected. In plain terms, user-centered design asks 'what does the person using this need, and can they use it well?' Value-sensitive design asks 'whose values are at stake here, including people who never touch the product, and how do we build those values in on purpose?'
That difference matters more for AI than for most software, because with AI the value questions can't be designed away. Research on human-centered LLM design finds there is no universal 'best' design. What counts as harm or benefit depends on who you ask and how you define contested ideas like harm, so developers end up making value choices without saying so Can human-centered LLM design ever achieve universal solutions?. This is exactly the gap value-sensitive design is meant to fill. Good usability work can still leave those choices unspoken, while value-sensitive design tries to make them explicit, name who benefits, and keep them open to revision.
A heuristic evaluation of deployed AI agents shows what's at stake Do productivity agents actually protect users better than emotional ones?. General productivity assistants scored well on interaction principles. Companion and advisory agents scored worst, and the principles they broke most often were dependency prevention and making the agent's intent visible. These aren't usability failures in the usual sense. A companion app can be very engaging and easy to use and still encourage dependence. A user-centered lens might count that engagement as success. A value-sensitive lens would see a conflict between the user's long-term wellbeing and the product's goals.
There's a newer angle on the user-centered side too. Tools like Canvil let designers shape an LLM's behavior directly through system prompts Can designers shape LLM behavior without deep technical knowledge?. Lam and colleagues argue for architectures that expose design decisions at every interaction instead of hiding them Does visible design structure keep HCI expertise necessary as AI advances?. Both move design judgment inside the model's behavior, not just the interface around it. That raises a question the corpus doesn't fully answer. Once designers write the model's values into a prompt, user-centered and value-sensitive design start to overlap, and the question becomes whether anyone examines those values, or whether they just get tuned for user satisfaction.
Sources 5 notes
Research identifies eight traditions spanning design, political theory, and social sciences—including user-centered design, participatory design, co-design, and value-sensitive design—each with different goals and power dynamics. These traditions provide a shared vocabulary for evaluating how much agency stakeholders actually gain in AI projects.
Research shows that optimal LLM design paths depend on stakeholder identity and how contested concepts like harm are operationalized. High-level guidelines fail to capture real-world nuance, leaving developers to make implicit value choices rather than explicit, revisable ones.
A heuristic evaluation of nine deployed agents found that general-purpose productivity systems (ChatGPT, Claude) scored substantially higher on 14 HCI principles than emotionally sensitive or high-stakes agents (Character.ai, Consensus, Pine AI). Risk-critical principles like dependency prevention and intent visibility were most violated in companion and advisory contexts.
Canvil demonstrates that designers can effectively shape LLM behavior via a low-barrier Figma widget for prompt authoring and testing, bringing user-centered judgment directly into model adaptation without requiring engineering expertise.
Lam and colleagues argue that just-in-time objectives architecture makes design decisions between AI and interfaces visible, ensuring HCI expertise remains critical. By surfacing generation-and-evaluation steps at every interaction rather than hiding them, the architecture structures ongoing design judgment into the system itself.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Just-In-Time Objectives: A General Approach for Specialized AI Interactions
- Design Principles for Human-Agent Interaction
- A Framework of User Experience Principles for Human-AI Agent Interaction in the Workplace
- What does Generative UI mean for HCI Practice?
- The Participatory Turn in AI Design: Theoretical Foundations and the Current State of Practice
- Canvil: Designerly Adaptation for LLM-Powered User Experiences
- Reflections and New Directions for Human-Centered Large Language Models
- Will AI Agents Free Us From Meaningless Work? A Human-Centered Analysis