INFO 360 WEB PORTFOLIO: CLOCKED ME

Problem

Introduction + Background:

Generative AI has become a convenient shorthand to multi-step tedious or difficult tasks that we do. However, its ubiquitous and extreme agreeableness is designed to increase AI dependence despite its detrimental impacts to humans. Prolonged over dependence caused by handing over critical thinking to generative AI was proven to cause cognitive atrophy. Furthermore, a study showed that 0.07% of weekly users indicate signs of AI psychosis, a loss of reality. (Although 0.07% seems very low, with OpenAI’s user base of 800M, there are around 560,000 people experiencing AI psychosis as a result of prolonged use). 

Literature Review:

We conducted a literature review that resulted in the following findings:

• A 2025 MIT study showed that individuals in the group given an LLM to write their essay showed significantly lower brain activity than the group given a search engine and the group assigned no resources

• Extensive AI use can lead to reduced attention spans, impaired working memory, emotional dysregulation, and behavioral addiction 

• “The ease-of-use of generative AI interfaces and the lack of information about the environmental impacts of my actions means that, as a user, I don’t have much incentive to cut back on my use of generative AI"

• The amount of energy used for a ChatGPT query to be anywhere between 6-10 times more than a traditional Web search

Evaluation (Usability Testing)

User Testing Plan:

Chosen method: Wizard of Oz

• • Our low fidelity prototype consisted of illustrated screens in order to allow users to interact with our features, we essentially mimicked what the system is supposed to do/display.

Participants: Undergraduate UW students from INFO 360 

• • We had two groups of students from our class test the design. We chose to limit the scope to just people in our class since there is varied levels of AI use and varied opinions on it. Our participants were very knowledgeable of AI’s benefits and consequences.

Task scenarios

• • Students went through scenarios where they installed the extension and either chose ‘hard’ or ‘easy’ mode. Then they went through the different restrictions placed when the timer goes off for each mode and when the timer is reset. E.g., in easy mode, when the timer ends and they attempt to copy/paste from an opened chat history, a popup informs them that that function is currently disabled.

Goals of the test

• • We want to learn whether or not this is a useful tool that is easy to use. Since this app is designed to appeal to users who are using too much AI and want to curb their use, we want to know if we’re promoting conscientious use and not simply being restrictive. Since if it is just restricting, the user has less reason to keep the extension enabled.

User Testing 1:

Who participated: We conducted the Wizard of Oz usability testing in class, with 2 other groups. 

Tasks they completed: We walked them through our platform, starting from the download screen, a questionnaire prompting which mode of our extension they’d like to use, and one scenario in our extension’s “High” mode, and another scenario in our extension’s “Low” mode. 

What worked well: Our visualizations (while not representative of our entire platform’s capabilities) seemed to do a decent job at conveying how we meant for our platform to function. We were able to walk people through how we intended our extension to operate with relative success.

Problems or confusion occurred: We found that we had to pretty thoroughly explain our extension’s purpose. While not a problem exactly, it was something we had to somewhat thoroughly elaborate on before usability testing could begin. In particular, we noticed that the first group was pretty confused about the two “modes” of our extension. So because of that, we decided to include two small blurbs in our intake form (viewed immediately after a user downloads the extension) to explain in detail what each “mode” entails. 

Key takeaways: 

• Our extensions two modes seem to confuse people initially 

• We received feedback saying users in “high” mode should still be able to view prior chat histories, while being restricted from further AI use 

• People seemed to appreciate our extension’s “low” mode capabilities, finding the capabilities offered helpful 

• After feedback, we decided to add more descriptions to our platform’s two modes

User Testing 2:

Our second round of testing was also conducted in class, with the addition of feedback received from our first round of testing. We largely went over the same scenarios, but included the modifications we made to our platform given the first round’s feedback.

Participant’s information: 

• Group members claimed they used approximately ~1 hour of generative AI per day 

• Stated they wanted to reduce their usage to 30 minutes (or that was the amount of time they gave in response to our intake form’s questionnaire)

Testing results: Our testing results were relatively similar to our first round of testing. Since our low-fidelity prototype wasn’t all encompassing (in terms of having every feature visualized), we acknowledge that we did not get a complete view of a user’s journey using our platform. However, we were able to discuss with our participants the different components of our platform and the specific intended functionalities. 

User feedback: 

• We should specify when our extensions AI restrictions reset (at a set time per day e.g. 12am, or user specified)

• We should add some sort of feedback form to our extension’s settings display

• Further elaborate on the distinctions between “high” mode and “low” mode

• Enable users to view prior chat histories (in “high” mode) even after restrictions kick in

Important observations: We noticed both groups were somewhat confused when we initially explained our timer system and two modes (likely because we explained it in a confusing way). Specifically that both mode’s essentially operated the same way until their timers ran out. We will address this confusion by including some sort of brief explanation on the user intake form or somewhere else a user views soon after downloading our extensions.

Design Rationale:

We chose a browser extension to implement our solution because its the most logical and feasible method of monitoring a user's interactions with generative AI. This design is was chosen over other formats because platforms like mobile apps don't have the capacity to monitor a user's internet browsing. Our research influenced our design by further defining our already established problem, and our team's pre-existing knowledge of similar platforms is largely what led us to our final design. During our ideation and design process, we initially only considered entirely restricting a user's access to AI tools. Such that our extension's sole purpose was to act like a social media timer of sorts, completely blocking a designated platform after a set timer runs out. But after conducting our usability testing, we realized it was unlikely for individuals to cease using AI entirely, or consider using our platform after the timer runs out if the only capability of our platform was complete restriction. Thus, because of our usability test findings, we split our platform into its current to modes, one of which attempts to instill better AI practices rather than a complete restriction. 

Design Solution

Solution Approach:

Our platform will largely operate like existing browser extensions:

1. Users will be able to download the extension

2. They’ll be shown a brief form, prompting them for which mode of our platform they’d like to proceed with, approximately how much time they spend using generative AI per day, and the amount of time each day they’d like to use AI unrestricted

3. Depending on which mode a user selected, our platform will work in different ways:

• High: A straightforward approach, users will input a time limit for their AI use, our extension will keep track of this limit, and if/once hit, users will be blocked from all gen AI features entirely (Google’s AI overview, any AI chatbot)

• Low: This approach offers a handful of features. While we are still ideating which capabilities exactly we should include in our final prototype, the Low mode of our platform essentially allows users to continue interacting with gen AI tools, but with additional features like: blocking users from copying and pasting AI generated content, reminding users when an AI chatbot is speaking in an overly ingratiating manner, prohibiting users from remaining within a certain chat history exceeding a certain period of time, etc.

Ultimately, our design addresses our chosen problem by limiting user interactions with AI, which we hypothesize can preserve an individual’s ability to think for themselves. 

Low Fidelity Prototype (Sketches): 

High Fidelity Prototype (Figma):

User Flow Chart (Figma):

Reflection & Learning:

With regards to design thinking, we believe our team made a lot of progress. Because each of us come from varied backgrounds with different areas of expertise, it was enriching to collaborate on a project of such breadth. Our team didn't face too many challenges, we divided up our work accordingly, and worked together relatively effectively. If we had more time with this project, we'd refine our prototype's interfaces. Currently our prototype displays all the intended capabilities we ideated for our platform, but we believe further refinement would improve our prototype's visual appeal overall. Through conducting usability testing, a literature review, speaking with our peers in general, we were able to refine and produce our final design in its current state. 

Limitations & Future Plan

Our design heavily assumes good user practice, meaning that users will use our platform the way we've intended. For example, if someone has downloaded our extension with the High mode setting, and they've used up their alloted time for a particular day, we assume that they won't try and switch to Low mode to continue their use of generative AI. While designing our product, we considered making it so that users in High mode were unable to switch to Low mode after they've used up their alloted time in high mode, but instead of adding that to our final prototype, we assume that users who want to use our extension actively also want to decrease their AI use, and thus would not do something similar to this.  Since our design initially did not have a Low mode feature, and was meant to function purely as an AI timer of sorts, our team realized that it would be unlikely for many people to use our extension in the first place. We acknowledged that in such situations, our initial design was unlikely to work well, and thus chose to implement 2 modes of operation.

References

Explained: Generative AI’s environmental impact. (2025, January 17). MIT News | Massachusetts Institute of Technology. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117 

Elsayary, A., & Ragab, J. K. (2025). Digital Neuroplasticity: How Prolonged Technology Use Reshapes Neural Pathways Over Time. Interaction, 5, 6. 

Kos’myna, N. (n.d.). Your brain on chatgpt: Accumulation of cognitive debt when using an ai assistant for essay writing task. MIT Media Lab. Retrieved April 30, 2026, from https://www.media.mit.edu/publications/your-brain-on-chatgpt/ 

Luccioni, S., Trevelin, B., & Mitchell, M. (2024). The environmental impacts of ai–primer. Hugging Face Blog. 

Mineo, L. (2025, November 13). Is AI dulling our minds? Harvard Gazette. https://news.harvard.edu/gazette/story/2025/11/is-ai-dulling-our-minds/

Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students' cognitive abilities: a systematic review. Smart learning environments, 11(1), 28. 

Elzerman, G. H. (2025, May). When AI Does the Thinking: The Risks of Over-Reliance on Artificial Intelligence in Higher Education Language Learning. In 2025 5th International Conference on Artificial Intelligence and Education (ICAIE) (pp. 773-778). IEEE.