AI-Powered Search UX
As an example website for this AI Search personal study project was taken an EUR-LEX website. EUR-LEX is the main EU Law repository with access to EU legal documents, available in all 24 EU’s official languages and updated daily.
It is assumed that the most used filters are:
Collection type (almost always used),
Document reference (used sometimes),
Dates (almost always used),
Languages (multi-selection is necessary),
and, of course, a text box for search.
And main users are: Experts, using the European Parliament website is part of their job
GOAL:
Come up with a better search that would include AI UX concept that would address the UX challenges outlined and respect constraints and design KPIs.
Standard Search Engine
A User will be able to choose from 2 possibilities of a Search engine:
Standard Search Engine - non AI-powered.
AI powered Search.
Firstly, let’s see a scenario when the User chooses the Standard Search. In this case, the User will be able to search by keywords and use filters. Filters presented are the most used/popular filters. By placing them near the search there is higher probability that the user will use them, so they will be able to narrow down results quicker, making the search more efficient and precise.
When clicking on search by keywords the User will see Popular searches and also searches the user have previously saved.
Important to note that I avoid automatic loading of a page after each filter. As it is disturbing and slows down the page. I prefer when a user can set all the filters they would like to and then trigger the load.
In case if a user is unsure which collection to choose they can choose a few at the same time.
For choosing a date I gave a few options to choose from: a user can choose a date range, scope, or a specific date. Which allows to narrow down the search results even further.
When listing various languages I avoid using country flags as for some users it can be complicated, also it might mislead to thinking of choosing a region instead.
Here a user can type the name of the language to find it quicker, as well as choose multiple languages.
AI Powered Search
When using the AI Powered Search a user can type a prompt. When start typing the AI search will suggest popular searches/documents on the topic being typed. Also a user can refine query by using filters on the side. There is a tooltip with explanation on what is AI search and how does it work.
AI search result:
Once the user submits a prompt to the AI, it processes the input and returns results with source and % of information taken from that source. Also, it is important to mention Confidence level for transparency and decision-making support. In order to keep improving the AI search it would be offered to vote, whether the result was helpful and met expectations or not.
Depending on the prompt - in this specific scenario a user requested key information, so the search offered them a summary and a full document below.
While reading the result, a right-click allows to quickly access further options and functionalities tailored to specific words/sentences highlighted, for example: copy the hihglighted text, making a prompt out of the highlighted area, save in the account, or broaden/elaborate on that part.
On the right side, there are further functions as: specify and elaborate on given result, for wider accessibility there is a function to read the result loud, option to download and to save to an account.