AI-powered Client Dashboard

I designed an LLM-powered web app Quologue, which pioneers a new way for users to interact with their e-book highlights.

MY ROLES

Research

Interview

Wireframing

Prototyping

UI Design

TITLE

Project Lead

Research Assistant

DATE

2021.06 - 2022.05

OUTCOME

TODO

DESIGN PROBLEM

blah

In this study, we examined the effects of physical appearance (more humanoid vs. less humanoid) and communication modality (Voice, Speech Bubble, and Objective Dashboard) on trust and perceived anthropomorphism.

Project summary

E-book users rarely re-read e-books, and they value e-books less than physical books. I look at e-book highlights as a potential resource to tackle this problem. The traces people leave on e-books have rich potentials for self-reflection and social interaction, but these are largely overlooked.

I designed Quologue, a GPT-integrated web application that allows users to reconnect with their e-book highlights through ongoing dialogue.

Experiment design

Task groupings

By performing median split on the data we gathered from Experiment 1, we assigned 6 descriptors to each of the ten warehouse tasks.

Results

Trust in robot was higher when 1) participants viewed a more humanoid robot, and 2) the type of task was Clean (Safe), Physically Easy, or Social. Also, participants who viewed a more humanoid robot reported higher levels of overall perceived anthropomorphism, Sociability, and Animacy.

No significant relationship was found between communication modality and trust in robot, but participants who viewed the robot with voice perceived the robot as having more agency compared to participants who viewed the robot with an objective dashboard.

My thesis paper lays out the specifics of my research process, theoretical background, and statistical analyses. If you're interested in checking it out, click here.