Background
About ChickPT
ChickPT is the largest part-time job platform in Taiwan, enabling employers to find excellent part-time talent and providing job seekers with various opportunities to earn and save money.
Project Brief
In the original resume photo upload process, an AI-powered evaluation feature was added to help determine if a profile picture is suitable and provide specific adjustment recommendations, such as suggesting to take a clear, front-facing headshot, so that users can easily make improvements.
Project Objective
Employers often value the professionalism of profile pictures when screening job applicants. The aim is to help job seekers understand how a suitable profile picture can increase their competitiveness in the job search process and assist them in making adjustments to improve their chances of success.
Persona
Young people aged 18-24
The largest group of job seekers, and the age group with less experience
Design Process
1
Feature Planning
Plan the flow and features
Confirm the feature content
2
Design Delivery
Create Hi-Fi designs
Create animations
Deliver Hi-Fi designs
3
Testing & Optimization
UI and UX testing
Refinement and adjustments
Intro of AI Features
As long as standards are set, AI is well-suited to perform repetitive and time-consuming tasks. This perfectly matches the need for profile picture evaluation. AI can quickly and accurately determine whether the clarity and composition of a photo meet the standards, significantly increasing efficiency and reducing labor costs.
Regarding specific standards (such as prohibiting non-real photos, excessive filters, and blurry photos), we used unqualified photo samples to train the model, allowing the AI to more flexibly handle various profile pictures.
The engineers also tested various models. For example, some models could achieve strict standards close to ID photos, while others focused on filtering out filter effects in photos. Ultimately, we selected a model that was accurate and moderately strict, enabling AI to ensure profile pictures meet the standards while also allowing users to adjust according to recommendations.
Goals
Design Goals
The operation should be simple and clear, avoiding user reluctance to use it due to AI being a new technology
Enable users to trust the prediction results
The AI operation process is a black box, and there is a need to reduce user doubts about its results
Not interfere with the original application process
The AI prediction cannot affect job seekers' application rights, ensuring fairness
Design
Goal: Reduce learning costs
In the early stages of the process, the existence of the AI prediction feature was explained, and the standard requirements for profile pictures were clearly listed to help users understand the basic rules before uploading photos, avoiding repeated operations that cause frustration.
In addition, the AI prediction is automated; users only need to upload a photo to receive immediate feedback. The operation is simple and intuitive, requiring no additional learning.
Goal: Enable users to trust the prediction results
To increase users' trust in the AI prediction results, I designed an analysis animation as a transition after the photo is submitted and before the results appear to alleviate waiting anxiety.
Furthermore, even if the internet speed is fast, the prediction results will not be displayed immediately. Instead, it is ensured that the animation plays for at least 3 seconds before the results are presented, based on the Doherty Threshold principle: Purposefully adding a delay to a process can actually increase its perceived value and instill a sense of trust, thereby avoiding user distrust of the results.
Image: Cover description and animation
Goal: Not interfere with the original application process
The AI prediction is a supplementary feature for uploading photos, aiming to help users optimize their photos rather than creating obstacles. Even if the prediction score is low, users should be able to complete the application smoothly. Therefore, only clear and practical suggestions are provided on the results page, and users can decide whether to accept our recommendations.
Iteration Proposal
Problem Description
From the data of job seekers being reported, it was found that some users with AI prediction scores of 0 or those who did not use the prediction feature had high activity, frequently applying and inquiring with employers, which created negative impressions on many employers.
This situation led employers to question the quality of talent on the platform, thereby affecting the overall credibility and image of the platform.
Image: High activity of reported job seekers
Proposal One: Add Alert to make the process more complex
Provide appropriate interference for users with low profile picture scores by adding a pop-up reminder in the two-step application process. By interrupting the process, prompt them to adjust their profile pictures to meet the expectations of the platform and employers.
Image: Comparison of application process before and after the change
Proposal Two: Add Profile Picture Border Design
Add a noticeable border and text prompts to low-scoring profile pictures to attract users to click and change their photos. This design is both guidance and a friendly reminder to encourage them to actively improve.
Image: Added prompts on the application page and resume page
Final Results: Overall Improvement in Profile Picture Quality
0-Score Ratio
-12 %
100-Score Ratio
+4 %
Learning Reflections
In this project, I learned which features are suitable for AI implementation and considered the relationship between users and AI, thinking about how to enhance user trust in AI. On the other hand, a data-driven design process helped me understand the gap between user behavior and needs, allowing me to make design adjustments based on reported data. Overall, balancing technical implementation and user experience, as well as the importance of data analysis for design strategy, were very important.
