Mosaic
A serendipitous search engine.
Duration
May to June 2024 (6 Weeks)
Team
Individual Project
Practice
Research, Market & UX research, UI design, Product thinking & Prototyping
Tools
Figma & Adobe Illustrator
Solution
By the end of six weeks and after extensive study of information seeking behaviors we were able to propose Mosaic - an new age search engine using AI to help increase chance encounters in search.
Mosaic as of now can be experienced as a set of flows imagined as possible browsing methods or SERPs.
“ DO NOT USE A.I. FOR YOUR RESAERCH!.”
This project, undertaken as part of a design exploration seminar at IDC, aimed to explore the impact of AI & LLMs on serendipitous discovery within digital search environments. By researching concept of serendipity, analyzing user search behaviors, and examining the deterministic nature of LLMs, the project explored how these technologies shape the user experience and how might we as designers make information seeking interactions serendipitous!
Primary Research
Since the study was done under design exploration semester, it was decided to ground the design of the process in the available literature. Hence to thoroughly understand the topic, literature study was divided into 3 sections.
Current Landscape
of web based information seeking A.I. Tools
The recent announcement of “ChatGPT search platform - SearchGPT” points to increased adoption of AI/LLM driven chatbot for information access or search. If this trend were to continue unchecked, precision and efficiency will become key measures of search or search engines. However, this emphasis on speed may come at the expense of serendipity.
Insights
Gathering from all the studies I attempted to make a browser that encourages Serendipitous encounters !
Low Fidelity Protoypes
Post peer discussion of 10 ideas depending upon factors like executability, time and appropriateness to the topic, 3 out of 10 ideas were prototyped. First low fidelity prototypes with interactions and later they were discussed with guides to be converted into concepts.
Prototypes
Concept -1
The project was titled - Mosaic . The name here represents a collection of items / information that comes together to showcase the whole image/overview of the topic in the query. Users are greeted with a main search page where users can input their query. After adding the query they are presented 2 options.
Goal search- is an iteration of focused search like current LLM chatbots, useful for situations where one needs fast and precise answers.
Omni search - focus on presenting users with a larger context of search query. It does that by showing keyword and related topics, that user can pick from before making the query, this here broadens the scope of search and introduces new avenues to show information that the user might find useful, making use of their context and planting keywords at a part of their journey where they can't ignore or miss them.
From here onwards Exploration 1 focuses on display of information in form of small windows arranged together where users can scroll horizontally and interact with information windows and each window also acts as a widget hence users can interact and engage with more information while being on the search page.
Concept - 2
Second iteration focuses on display of information, here the omni search would take user to a wheel and spoke like display of related terms that the user can go through and then edit and change, here also the idea is to delay the interface with the information itself and first broaden the search area and prepare user so as to when information is presented they can see some value in that. These spokes can be increased with a slider on the right side of the screen and users can also edit these spokes then to create new context for their search.
Once the user is satisfied with the terms they can also explore the information that can be found on the intersection of the term at the center and at the periphery. Here, instead of showing all the information about the query and related terms, the exploration focuses on the information that is at the intersection of the 2 queries by the use of LLMs. These links or information can also be saved inside the canvas for a more contextualized search and then can be revisited also.
Concept - 3
The third iteration focuses on display of information once the user has selected Omni search, here in display of information, a larger overview of information/ query is presented with help of LLMs first to the user, this in turn then first focuses on familiarizing the user with the topic and its complimentary topics before looking into a much more detailed view. With this view users can develop an overview of the topic and then explore or make unexpected connections. Moreover this also enables users to look into directions that an instant detailed answer would have left no scope for. The prototype does so by establishing a branched tree like structure from top to bottom, where the topic/query is broken down into different aspects and then further broken down to smaller levels last to contributors or stakeholders of the space.
Thank you !
Let me know if you’d like a detailed copy of the project.
yashrawt458@gmail.com