As technologies evolve and AI-powered products become more present, UX designers encounter new types of challenges. Data introduces a new layer to the product experience for users. Thus, in addition to collaborating with software developers, designers now also need to work with data scientists. Based on my academic and industry research, I will share insights about this type of collaboration. I will also explore the differences in the daily workflow with data scientists and software developers, highlighting key skills designers need to learn to increase their effectiveness in this process and shape better products.
Watch this WaysConf session, then continue with related talks or explore the current programme.
In her WaysConf 2024 presentation, Anna Maria Szlachta addresses the AI innovation gap caused by AI products failing to solve the right user problems. She argues that bridging this gap requires deep, structured collaboration between product designers and data scientists. Introducing the concept of User Algorithmic Experience (UAX), Szlachta breaks down AI products into three layers: data, model, and interface. This framework shifts the ownership of user experience from being a purely front-end concern to a shared responsibility with data scientists. To foster this collaboration, designers must learn basic data science terminology, involve data scientists directly in user research, and run collaborative workshops. Furthermore, Szlachta highlights her academic research on using visual language to help non-technical users articulate their experiences with abstract algorithms, which dramatically increased user feedback quality. By establishing continuous user-model feedback loops and using gamification, product teams can build more realistic, context-aware AI solutions that go far deeper than surface-level interface design.
The User Algorithmic Experience consists of three layers: data, the model, and the interface. Framing UX this way encourages data scientists to take ownership of the user experience rather than viewing it as a front-end interface issue.
Watch from 2:35Designers do not need to master complex mathematics, but they must learn basic data science concepts like accuracy, uncertainty, and true positives or negatives. This shared vocabulary allows both roles to effectively discuss model limitations and capabilities.
Watch from 3:51Data scientists should be actively invited to user research presentations because they are ultimately trying to model reality. Understanding real-world user behavior directly influences how they approach algorithmic problem-solving.
Watch from 5:13Using visual diagrams to explain underlying algorithmic mechanisms helps non-technical users articulate their experiences during research. In studies, this visual approach extended user interviews from 15 minutes to 1.5 hours and revealed deep, previously unmentioned insights.
Watch from 9:43Users are often unaware that their actions influence AI models, making it crucial to design clear feedback mechanisms. Implementing gamification and targeted prompts helps gather the specific human validation data that scientists need to improve model quality.
Watch from 11:35These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.
Despite the massive hype surrounding generative AI and machine learning, many organizations still struggle with what specialists call the AI innovation gap. This issue persists because development remains highly technology-driven, leading to products that fail to solve actual user problems. The root of this challenge often lies in the poor collaboration between product designers, user researchers, and the data scientists or engineers building the underlying algorithms.
Watch from 1:00To bridge the gap between design and engineering, it helps to introduce the concept of User Algorithmic Experience, which is split into three distinct layers: data, the model, and the interface. When designers frame the product this way, data scientists stop viewing user experience as a superficial front-end concern. Instead, they realize that the data and models they build directly shape the user's journey, fostering a shared sense of ownership.
Watch from 2:35Designers do not need to master complex mathematics, but they must learn the basic vocabulary of data science to communicate effectively. Understanding concepts like uncertainty, accuracy, and true positives or negatives allows designers to grasp the limits and capabilities of different models. This shared language prevents misunderstandings and makes it easier to dissect complex, multi-model systems together.
Watch from 3:51Traditional interview methods often fail when researching algorithmic experiences because non-technical users struggle to understand abstract backend mechanisms. By introducing visual diagrams that illustrate how the algorithm works behind the scenes, researchers can unlock incredibly deep insights. In experimental studies, showing these visuals extended user interviews from a brief fifteen minutes to an hour and a half, enabling participants to connect their real-world experiences to the system's behavior.
Watch from 9:43AI systems require continuous learning, but users are rarely aware that their behavior directly trains and influences the model. While data scientists initially ask for as much data as possible, designers can help them narrow down specific goals and numbers. By implementing gamification and clear interface elements, designers can encourage users to validate predictions, creating a high-quality feedback loop that improves the model over time.
Watch from 11:35
