Empathy Map demonstrates that residents see potential for GenAI to be a fun experience.Residents need fun ways to build confidence
Residents who have positive or neutral assumptions about GenAI prefer self-exploration of the tool.
Preference towards conventional sources limits the willingness to rely on GenAI for critical information.
In cases where UM-GPT provides a non-answer, participants start to expect it to give vague responses for future uses.
Participants found conversing with UM-GPT to be entertaining, but were still skeptical and preferred familiar sources like Google or their personal network.
Journey Map highlights that residents feel anxiety when choosing between different resources to get their questions answered
Responses to the survey question: Which of the following is the most important when getting your question answered about the City of Ann Arbor?2. Residents feel anxious about new technologies due to a lack of understanding
Our research demonstrated familiarity is important to users when choosing an information source, especially when the user has feelings of uncertainty or anxiety. Because of this, our team suggests implementing and advertising opportunities for less experienced users to familiarize themselves with GenAI. Some examples of ways to do this are listed below.
Educational Workshops: Host workshops at the library to showcase GenAI’s capabilities. These sessions can provide hands-on experience and demonstrate practical applications of the technology.
Webinars and Community Events: Set up informational booths at city council meetings or community events to teach residents about GenAI. Virtual webinars can also reach a wider audience without incurring significant costs.
University Partnerships: Collaborate with the Artificial Intelligence Lab at the University of Michigan to offer seminars and workshops. Leveraging academic partnerships can provide expert knowledge and credibility.
Website Enhancements: Include a prompt on the GenAI website that explains where the information comes from, along with a dedicated web page offering detailed information about the tool. This will help increase user literacy, trust, and familiarity in GenAI.
Empathy Map demonstrates that GenAI is not mirroring current search methods, which leads to resident dissatisfaction
Responses to the survey question: When you have a question about the City of Ann Arbor, how do you get that question answered?3. Residents are looking for genai responses to match what they are used to
Through our survey we found that participants are using the city website and an internet search the most currently, these methods must satisfy their need of receiving specific answers best. (see chart below)
Because participants’s feel overwhelmed with the length of the answers received from the GenAI interface, it would be beneficial to make sure the GenAI interface includes information hierarchy allowing residents to skim through for most relevant information. This can be done through bolded words, bulleted lists, structured headings, etc. This, in turn, would shorten the time taken by the residents to absorb and process information because the structure is easy to follow. Additionally, there is another opportunity to introduce this tool alongside educational resources to help branch the learning curve that residents will experience while using this tool. This can be done through providing tutorials or other educational resources such as videos or articles.
Responses to the survey question: Describe your feelings about using GenAI to answer questions about the City of Ann Arbor?4. Residents feel more confident in the GenAI when sources are provided
From our survey, we asked residents to "Describe your feelings about using GenAI to answer questions about the City of Ann Arbor." We organized approximately 50% of the responses into categories as illustrated in the chart to the right.
Recommend enhancing the transparency of information produced by GenAI by explicitly marking all disseminated information as originating exclusively from the City of Ann Arbor's website. To this end, GenAI could include explicit hyperlinks directing users to the exact locations on the city website from which it sourced its information. By making this clear, residents will understand they are accessing consistent information through the interface, regardless of whether they perform their search via the government’s website or contact the City directly. We believe that this level of transparency will significantly bolster residents' trust and confidence in the precision and dependability of data provided via the GenAI interface
Responses to the survey question: Describe your feelings about using GenAI to answer questions about the City of Ann Arbor?5. Residents need prompting to enter follow-up questions to get higher answer satisfaction
The fifth and final insight determined is residents need prompting to enter follow-up questions to get higher answer satisfaction. 92% of residents (from the Moderated Field Study) do not know or use follow up questions in their interaction with GenAI. It is important to recognize that a lot of people do not know or do not use follow up questions to refine their question and deem the responses by GenAI inaccurate and/or unhelpful when GenAI potentially just needs some redirecting and refining of the question.
In terms of a recommendation that follows this insight, we suggest that the City of Ann Arbor implement something that prompts people to ask follow up questions for further clarifications and refined answers. By doing so, GenAI can increase confidence and satisfaction of users by answering the question in a more specific way