NaRa (short for narrative) is a research project at Harvard University that explores what AI might look like when it behaves less like an authoritative answer engine and more like a carefully restrained self-reflection object—one that supports people in thinking without taking ownership of their thoughts.

The project began with a concern about how young people (18 - 25) increasingly turn to conversational AI for questions that are not merely informational: Should I leave this relationship? Am I making the right choice? What does this situation say about me? In these sensitive moments, answers can become a form of cognitive and emotional offloading. 

On top of that, the popular LLM interfaces we mainly engage with today have been designed to be reassuring to the user as a baseline feature. Data shows that AI endorses the users' positions 49% more often than humans.

NaRa does not ask users to choose between using AI and rejecting it.The problem is not that AI may be wrong. It is that a confident response can prematurely close down ambiguity, reduce critical reflection, and weaken the user’s sense of authorship over decisions that should remain personally theirs. 

Inspired by LEGO SERIOUS PLAY 

One of NaRa's sources of inspiration has been the LEGO SERIOUS PLAY method: 

"LEGO SERIOUS PLAY has done something we deeply admire by using the hands and a physical artefact to unlock thinking and relational awareness that conversation alone cannot reach. We are trying to do something adjacent with our project, but in the context of AI-saturated environments. We're seeing young people are increasingly offloading their emotional processing to systems that validate rather than develop them. We are particularly studying the boundary between play and learning, specifically the moments in LEGO SERIOUS PLAY workshops where something shifts from performed engagement to genuine insight."

NaRa is a compact physical interface that lives alongside the phone, supported by a browser extension and web application. NaRa’s approach resists the idea that technology (AI) should supply the “correct” interpretation or answer to every question. Instead, meaning is constructed by the participant, allowing tacit knowledge, tensions and relationality to become visible and discussable. NaRa externalizes a question through symbolic fragments (glyphs) that must be interpreted rather than consumed.

NaRa does not generate a paragraph of recommendations. It responds with only one word and three ambiguous glyphs. These cues are contextually related to what the user has expressed, but they are intentionally incomplete. Their purpose is not to tell the user what to think, but to create enough distance and friction for interpretation, inner dialogue, and individual sense-making. The glyphs are not icons with predetermined definitions. They act as provocations that help users articulate something they may already sense but have not yet formed into language. NaRa’s design philosophy, integrated to the interaction and user experience, benefits from metaphors, ambiguity, and the movement between seeing and contextualising. 

For more information about the project or watch video
To connect with the project team send email

Sincerely,