Year
2025
Role
Lead designer and researcher
Context
IIT Hyderabad
Duration
Sep 2024 to Apr 2025
Team
Solo project
Status
Concept, tested with ten interviewees
The problem
Young adults feel isolated even while constantly connected, and today’s assistants help with tasks, not with the nerves of a new place or a room full of strangers.
The outcome
ARiA, a chibi-faced companion for phone and mixed reality, with six modes for everyday moments, each one built on a finding from the review and the field research.
42.3%
felt isolated often or always
Survey, 46 young adults
70%
avoided asking for help in new places
Field observation, 10 people
82%
named privacy as their top concern
Same survey
Overview
This project picks up where my literature review of mixed-reality companions ended. The review found four gaps; ARiA designs into them. The goal: a companion people would actually trust in ordinary moments, like a new campus, a crowded event or a tense evening, that encourages real connection instead of replacing it.
Concept render: ARiA points the way to a metro exit.
DISCOVER
Challenge
I surveyed 46 young adults and watched 10 people on campuses, in cafes and at gatherings. Most avoided eye contact, used their phone as a shield and hesitated to start conversations. They wanted an assistant that was empathetic and could read their mood, but trust was low: only 19.2% were definitely interested.
What 46 young adults wanted from an assistant, and what they feared.
Let’s understand this with a scenario
“The edge of the room”
Picture a student at her first college event. She stands near the wall, phone in hand, scrolling nothing. She wants to join the group by the window but has no idea how to start, so she leaves early. She didn’t need a lecture or a chatbot. She needed a small, friendly nudge at the right moment.
She wanted in. She didn’t know how to start.
DEFINE
How might a companion give people that nudge to connect, without feeling like surveillance or a replacement for real people?
People wanted warmth but feared being watched. So the companion had to earn trust: a friendly character, gentle gestures instead of constant talk, and restraint.
DEVELOP
Approach
I worked alone, from research to character. The review showed realistic human faces feel uncanny, so I dropped a realistic avatar for a chibi character: expressive enough to connect with, abstract enough not to unsettle. It went from hand sketches to a character sheet to a 3D model. Because the review found body language underused, ARiA leads with gestures and expressions, not voice.
Each of the six modes answers something seen in the research: Smart Navigation, Calm Mode, Social Spark, Emotional Support, Study Buddy and Women’s Safety. A concept video then showed all six across six everyday scenarios.
What I saw in the field, and the part of ARiA each behaviour shaped.
DELIVER
Result
In ten in-depth interviews after the concept video, the character did what a floating interface couldn’t: skeptics warmed to it. People asked for exactly the moments the modes target. One said: “If an XR assistant could give me conversation starters or calm me down before a social event, I’d definitely use it.”
The output is a research-backed concept: the character system, six modes, information architecture, wireframes and a high-fidelity mixed-reality interface.
ARiA’s character sheet: the expressions and poses it uses instead of words.
Key takeaway
Believing in a concept is not evidence for it. The character worked because research shaped it, not because I liked it.
Reflection
ARiA is a concept, not a validated product: no prototype was tested in the wild and no controlled study was run. A planned collaboration with a psychologist did not happen, which narrowed the scope. Next time I’d run the survey and observation alongside the review instead of after it. Next steps: a working mobile prototype and a long-term study.
Grounded in
Bhagat and Remedios, The Potential of Mixed Reality (MR) Companions for Navigation: A Systematic Literature Review (ICWIHI 2025, Springer)
The review this design responds to.
