For decades, principles relating to usability, predictability, and efficiency have guided UX design, ensuring that users can complete their tasks with minimal cognitive effort. With traditional, direct-manipulation graphic user interfaces (GUIs), users take action, then the system responds in consistent, expected ways.
However, generative artificial intelligence (GenAI) disrupts this UX design paradigm by introducing an interaction model that is based on the specification of intent-based outcomes. [1] Instead of following predefined workflows, users describe what they want, and the AI generates variable results—often with unpredictable or emergent outcomes. [2, 3, 4]
This UX design paradigm shift that we’re currently experiencing challenges long-standing UX design heuristics such as consistency, predictability, and seamlessness. Designing for GenAI requires a new approach—one that embraces transparency, adaptability, and user control, empowering users to navigate uncertainty, iteration, and co-creation rather than expecting deterministic outputs.
Champion Advertisement
Continue Reading…
Why Traditional UX Design Falls Short in the GenAI Era
Two foundational principles have long guided traditional UX design:
predictability—Users expect the same result in response to the same input, ensuring consistency and reliability.
seamlessness—User interfaces should minimize cognitive load, making interactions feel effortless.
However, GenAI disrupts these assumptions, requiring a fundamental shift in how we design user experiences, as follows:
generative variability—Unlike deterministic software, GenAI can produce multiple, unpredictable outputs in response to the same input, requiring users to navigate variation rather than expect consistency. [5]
active engagement over passive efficiency—Instead of frictionless interactions, GenAI demands deliberate engagement, requiring users to experiment, iterate, and refine outputs to get meaningful results. [6]
2 Key UX Design Paradigm Shifts
GenAI requires two key UX design paradigm shifts, as follows:
From predictability to generative variability and exploration
From seamlessness to active engagement
1. From Predictability to Generative Variability and Exploration
Traditional software delivers consistent, repeatable results, but GenAI creates multiple possible outputs in response to the same prompt. [7] Instead of designing a single correct response, UX designers should create user experiences that support exploration, iteration, and refinement.
Example 1—Google Bard presents multiple draft responses to a query, reinforcing the reality that AI-generated text is inherently variable.
Example 2—MidJourney and DALL-E generate multiple image options for each prompt, shifting the user’s role from receiver to curator.
2. From Seamlessness to Active Engagement
Traditional UX design aims to streamline workflows and automate decision-making, but GenAI shifts the focus from efficiency to collaboration. AI is not an infallible expert, but a co-creator that requires user input, verification, and adjustment. [8]
Example 1—Adobe Photoshop’s Generative Fill lets designers modify AI-generated images using a familiar user interface, treating AI as an assistive tool rather than a fully autonomous system.
Example 2—GitHub Copilot suggests multiple code completions, but developers must actively review, test, and refine its suggestions, highlighting the importance of human oversight.
This shift from automation to collaboration requires rethinking user interactions, embracing uncertainty, iteration, and controlled friction rather than aiming for perfect predictability. A talk that Hayley Mortin and Di Le presented at UXC23, “From Jackpots to Algorithms: The Role of Human-Centered Design in Navigating Unpredictability,” explored similar themes, highlighting how randomness and intentional friction can enhance user experiences rather than hinder them. [9] Just as users must engage, refine, and interpret AI-generated content, they must also develop new skills to navigate generative variability rather than expecting seamless, deterministic automation.
Rethinking UX for GenAI: A Framework for Product Teams
Because GenAI disrupts traditional UX design principles, UX designers must prioritize transparency, control, and collaboration.
The framework that Weisz et al. presented in their paper “Design Principles for Generative AI Applications” provides a structured approach that is based on six key design principles. [10]
1. Design Responsibly
Ensure that AI is ethical, fair, and minimizes risks such as bias, misinformation, and copyright violations. For example, the maker of DALL-E discloses that they train the Generative Fill feature on stock and public-domain data.
2. Design for Generative Variability
Help users navigate AI’s unpredictability by offering multiple outputs and tracking version history. For example, MidJourney generates multiple image variations that users can refine, then select the best one.
3. Design for Mental Models
Guide users in understanding how AI works and adapting their interactions accordingly. For example, GitHub Copilot follows an autocomplete pattern, making AI-powered coding suggestions easier to use.
4. Design for Co-Creation
Give users more control over AI-generated content, making AI a creative partner rather than an automated tool. For example, Adobe Firefly lets users tweak AI-generated designs, ensuring that they remain active participants in the creative process.
5. Design for Appropriate Trust and Reliance
Help users calibrate their trust in AI by making its strengths, limitations, and information sources transparent. For example, ChatGPT warns users that its responses might be inaccurate, reinforcing users’ healthy skepticism.
6. Design for Imperfection
Prepare users for flawed AI outputs and give them tools to refine, regenerate, or give feedback. For example, Google Bard lets users modify AI-generated text—making it shorter, longer, or simpler—encouraging iteration.
Conclusion: The Future of UX in a Generative World
This is the final part of our six-part series on UX research for GenAI. Throughout our series, we have explored how research must evolve to meet the challenges and opportunities that GenAI presents. From reframing research approaches to understanding how GenAI impacts the user experience, we have examined the roles of trust, transparency, and human-centered design in shaping the future of AI-powered products.
As Weisz et al. emphasized, “Calibrating users’ trust is crucial for establishing appropriate reliance.” [11] The future of UX for GenAI is not just about making AI easier to use, but about ensuring that AI works with users, fostering transparency, exploration, and informed decision-making. As generative technologies continue to evolve, UX professionals can play a critical role in shaping AI-driven experiences that empower rather than overwhelm users.
[2] Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. “Design Principles for Generative AI Applications.” In Proceedings of the CHI Conference on Human Factors in Computing Systems, January 2024, arKiv.
[5] Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. “Design Principles for Generative AI Applications.” In Proceedings of the CHI Conference on Human Factors in Computing Systems, January 2024, arKiv.
[7] Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. “Design Principles for Generative AI Applications.” In Proceedings of the CHI Conference on Human Factors in Computing Systems, January 2024, arKiv.
[10] Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. “Design Principles for Generative AI Applications.” In Proceedings of the CHI Conference on Human Factors in Computing Systems, January 2024, arKiv.
[11] Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. “Design Principles for Generative AI Applications.” In Proceedings of the CHI Conference on Human Factors in Computing Systems, January 2024, arKiv.
After graduating with a Master’s in Experimental Psychology and publishing in the field of psychology and law, Katie began her UX career at Northrop Grumman where she was a lead UX researcher for enterprise experiences. She helped form the first team for enterprise UX at the company, then went on to manage several cross-functional teams focusing on internal and external products and experiences. Katie joined ServiceNow in 2022 as the manager for the Artificial Intelligence/Machine Language (AI/ML) UX Research team. Under her leadership, the team has grown in size and business influence, participating in history-making product roll outs for Generative AI. Her team has also emerged as a strong voice for the role of UX research in responsible AI and human-centered AI ethics. Katie values transparency, human connection, and loyalty as both a people leader and a voice in the field of AI. Read More
As a Senior UX Researcher on the Platform Artificial Intelligence/Machine Language (AI/ML) team at ServiceNow, Hayley started her journey in AI working as a Data Annotator, where she learned about the AI development lifecycle while creating datasets for training computer vision. This foundational experience paved the way for her transition into UX Research, a move that was inspired by her academic background in Psychology and Behavioral Science. Today, she focuses on understanding how users perceive and approach adopting AI/ML technologies, and she explores ways to build trust with users through explainable AI design. Read More
Manager & Strategist of AI UX Research at ServiceNow
San Diego, California, USA
Jessa has over 15 years of experience researching human behaviors and needs, with a PhD in Health & Human Behaviors, and nearly five years focusing specifically on helping to understand the user experience of artificial intelligence (AI) in enterprise settings. She is a champion for humans and elevating the role of the human in the unique interplay between AI technology and users, across a variety of personas, from non-technical to highly technical. Outside of work, she is busy being a mom and soaking up the sun in San Diego. Read More
