Edujob's first AI — Artificial Intelligence — was designed to help young people send a photo that would be clearer to the recruiter during the recruitment process.
The development and data team conducted research and surveyed the characteristics of the most frequently sent photos on the platform. Four categories of photos were found:
2 Project Manager
2 UX Researchers/Service Designers
2 Developers
1 CX Manager
After the accuracy of the AI was measured and we were ready to put it into production, many questions arose:
How could we inform the young person that the photo sent may not help in their recognition during the recruitment and selection process?
To better understand how the user's path would work, we structured a flow, with the four possible unhappy paths determined by the four classes, and a happy path where the photo would not be classified.
So we had to inform the user if it was classified as: photo of documents, photos of photographs, non-human photos and side photos. How frustrating for him, right? Unhappy path.
We would also have to inform him if the photo was not classified. Great, everything worked out. Happy path!
As the goal was to help young people send clearer photos to recruiters, we would provide information on exactly how the photo was classified.
In the flow above, we have already mapped out a possible message of how it would be to inform the user if the photo sent was classified as a photo of a document.
“Oops! This photo may not be clear to our recruiters. Our Artificial Intelligence has identified that you sent a photo that was taken of a document.”
People search for more information about the medics in the medical network search to make a decision
To understand how the young person would react to the classified photo, we mapped out a flow where we would receive feedback from the user about the evaluation made by the AI classification.
We thought of many paths and many alternatives, but we had to put the AI into operation to validate it quickly, so we decided to simplify the flows and some functionalities.
The messages about the classifications would be unique: through a carousel informing our classes along with other guidelines for sending a more assertive photo. We thought about the scalability of several scenarios that we could have in the future, and we decided to simplify at this time.
We decided to remove the young person's evaluation in relation to the AI classification, as we believe it would be an additional task at a time when he is still trying to understand what happened when he sent his photo, the intention here was to make the photo be resent.
The task success was to make the user send a more assertive photo if it was classified, so based on that we defined a metric:
Number of photos that were sent more assertively after being classified.
Consequently, number of more assertive photos in the database after the AI was working on the platform.
Utilizamos cookies para analisar o tráfego do site Web e otimizar a sua experiência de utilização do mesmo. Ao aceitar a nossa utilização de cookies, os seus dados serão agregados com os de todos os outros utilizadores.