How UX Practitioners Communicate AI Concepts: Insights from Hands-on Design Experiences

As artificial intelligence becomes more prevalent in user-facing technologies, UX practitioners (UXPs) face new challenges in designing and communicating AI-enabled interfaces. This study examines how UXPs communicate AI concepts after gaining hands-on experience with AI model training using Google's Teachable Machine tool. Through analyzing design presentations and interviews from 27 UXPs, researchers identified key themes in how UXPs work with and communicate AI as a design material.

UXPs reported significant challenges in collaborating effectively with AI engineering teams. Key issues included: - Work occurring independently and linearly between UX and AI teams - UXPs often brought in late in the development process - Lack of understanding about AI capabilities and limitations - Erosion of trust between UX and AI teams due to communication gaps - Difficulty 'bridging the gap' between UX and AI domains These challenges highlight the need for better communication and collaboration strategies between UX and AI teams.

When creating design presentations, UXPs emphasized several key aspects of AI communication: - Model selection rationale, comparing pros and cons of different models - Customer value and business benefits of the AI solution - Engineering costs and implementation considerations - Model performance, with a strong focus on accuracy Many UXPs considered accuracy the most critical factor to communicate, seeing it as essential for meeting user needs and driving discussions on how to improve the AI system. However, UXPs often struggled to effectively communicate technical aspects of model performance.

Using Teachable Machine to experiment with AI models had a significant impact on how UXPs approached AI communication: - Increased confidence in discussing AI capabilities and limitations - Better understanding of data quality issues and their impact on model performance - More concrete ideas for iterating and improving AI models - Enhanced ability to bridge communication gaps with technical stakeholders This hands-on experience helped UXPs develop a more nuanced understanding of AI as a design material.

UXPs demonstrated awareness of both the potential benefits and risks of incorporating AI into their designs. Key considerations included: - Ethical implications of AI-driven decision making - Privacy concerns related to data collection and usage - Potential for AI bias and its impact on users - Balancing automation with user control and agency Many UXPs incorporated these considerations into their design presentations, showcasing a holistic approach to AI-enabled UX design.

After experimenting with AI models, UXPs were able to formulate more concrete next steps for AI development in their projects: - Suggestions for expanding and diversifying training data - Ideas for refining model architectures and improving accuracy - Proposals for user testing to validate AI-driven features - Plans for iterative improvement based on real-world usage data This ability to propose actionable next steps demonstrates the value of hands-on AI experience for UXPs in driving project direction.

This study highlights the importance of providing UXPs with hands-on AI experience to enhance their ability to communicate and collaborate on AI-enabled projects. By bridging the gap between UX and AI domains, organizations can foster more effective interdisciplinary teamwork and create better AI-driven user experiences. Future research should explore how to integrate tools like Teachable Machine into UX workflows and develop best practices for AI communication in design presentations.