Enhancing user trust, decision confidence, and adoption of advanced models through guided selection

<aside> ⚠️

Disclaimer: Metrics are simulated based on observed user behavior patterns and used to illustrate potential impact.

Tool used: ChatGPT, Hotjar, Dovetail, Figma Make, Figma Slide

</aside>

Overview

Many ChatGPT users struggle to select the right model (e.g., GPT-4o vs GPT-3.5) for their task.

→ This leads to misaligned outputs, wasted effort, and reduced trust in the tool — even when the “wrong” result is due to model choice, not model quality.

<aside>

The goal: Redesign the model selection flow to make it intuitive, educational, and adaptable, increasing correct first-attempt selections and advanced model adoption.

</aside>

Problem

• Model selection is hidden or unclear, especially for new users.
• Users often default to GPT-3.5 simply because it’s already selected.
• Model differences are poorly understood until after output is generated.

Impact on users: wasted time, frustration, and loss of trust.

Impact on product: lower engagement with advanced models (GPT-4o) and reduced perception of value.

Research & Key Findings (AI-simulated metrics based on observed patterns)

• Low awareness – 60% of users didn’t know they could choose a model.
• Default bias – 70% defaulted to GPT-3.5 regardless of task type.
• Rework after mismatch – 45% of sessions required rewriting or re-asking because the initial output didn’t match expectations.

Reality Check: OpenAI Research (How people use ChatGPT)