General Audience: AI Illustration Quality Comparison

We're running a paid study to benchmark the quality of AI-generated illustrations. By gathering human preferences on image outputs, we can better align our models with what looks natural and accurate to a real person. This work directly feeds into improving how well the system follows text prompts and produces clean, polished visuals.

You will complete a remote session lasting about 30 minutes where you view 100 or more image matchups. For each matchup, you will read a short text prompt and look at two AI-generated illustrations side by side. Your task is to pick the better image based on how well it matches the prompt and its overall visual quality, choosing between left, right, tie, or both bad. You will also occasionally encounter clear calibration checks to ensure you are actively paying attention to the details.

We welcome general participants from all professional backgrounds to share their honest, gut-level preferences. No prior art or design experience is required to participate in this study. You simply need to be detail-oriented, comfortable reading short prompts, and capable of making quick, accurate judgments on visual quality.

• Review side-by-side matchups of AI-generated illustrations based on a provided text prompt.
• Evaluate each image for prompt accuracy, clean shapes, composition, and lack of visual glitches.
• Select your preferred image or indicate if there is a tie or if both are unacceptable.
• Identify and pass obvious calibration checks to confirm your active attention to the task.
• Complete 100 or more comparisons during an approximately 30-minute session.

• No professional art or design experience is needed
• Strong attention to detail to spot visual glitches, proportions, and prompt alignment
• Comfortable working through repetitive visual tasks for about 30 minutes
• Access to a desktop or laptop computer to view side-by-side images clearly

• You will be engaged as an independent contractor.
• This is a fully remote opportunity that can be completed on your own schedule.
• Opportunities can be extended, shortened, or concluded early depending on needs and performance.
• Your participation will not involve access to confidential or proprietary information from any employer, client, or institution.
• Payments are processed weekly based on services rendered.
• We are unable to support H1-B or STEM OPT candidates at this time.