Inara - Establishing Trust in AI
Following the design thinking framework, the empathise phase of the study consisted of semi-structured interviews (n=2) competitive analysis and a survey (n=34) to help inform the approach to the study and gain a better understanding of ethical shopping practices and trust in AI. Throughout the study all findings were documented and illustrated on Figjam
Problems identified
Semi-structured interviews and 15% of survey responses show concern around the ethics of usign AI
47% of respondents to the survey would like to shop more ethically but do not know what sources to trust.
15% of respondents would use AI as a starting point to inform product purchases but would follow up with their own research.
Results
By using Ecosia as the integrated AI and informing users of its ethics during onboarding trust in AI was established through transparency early on. Allowing users to rank their own ethical priorities gave them control over those results, reinforcing trust further.
The ability to verify claims through visible sources was a consistent measure of trust in both unmoderated and moderated testing.
60% of participants felt Inara gave them enough information to make a confident purchasing decision.
"After completing the onboarding, I have a clear understanding of what the app does (n=23)"
84% - Strongly Agree
Process - Design Thinking
01
Empathise
I conducted semi structured interviews (n=2), survey (n=34) and competitive analysis to understand the problem space. At this stage I created a proto persona that was informed by assumptions
02
Define
Wrote a How Might We and a problem statement grounded in the research, which set the direction for both the study and the design. Defining design principles up front kept the user at the centre of every decision.
03
Ideate
Wireframed a first draft to establish the flow and core patterns. A task flow bridged the wireframes and high-fidelity work, mapping the flow for the different screens and mapping these with existing design patterns for AI chat and scanning.
04
Prototype
Built an advanced Figma prototype with variables, so user input toggled layer visibility giving the feel of a real, working app the participant controlled. Designed on Material Design for its WCAG-aligned accessibility.
05
Test
Tested with the target segment in Useberry, capturing time on task, heatmaps, task completion, SUS, and post-task questions. Moderated sessions ran on a think-aloud protocol to pull sharper insight from participants.
“Seeing the ethics of Inara would make me trust using AI a bit more…without the sources I wouldn't trust the result ”
Participant 2
Skeptical AI user - Actively Ethical Consumer
Conclusion
When grading customer product choices there is an expectation that there is full visibility on how the app came to its rating, although sources helped develop trust, in future iterations more granularity would be explored.
Users believed Inara would be useful on a scan by scan basis when the user is curious about a product, adding to report to compile a overall rating is an action that did not follow existing mental models and testing proved that it did not have value and would be dropped if there were another iteration.