Artificial Intelligence (AI) in combination with user experience (UX) adds additional quality to the classic UX design. But does AI really request a special UX extension? Doesn’t a working UX already cover all needs including AI? Let’s conduct a detailed investigation.
SAP’s AI Design practice team carried out multiple user research activities on AI supported use-cases on AI features and AI patterns, which revealed three key qualities:
- Control: If they can control it, users will accept a supportive AI.
- Transparency: Users need to get the full information about the supporting AI service.
- Explanation: At a certain stage, users demand explanations about AI.They want to know how and why they receive pre-selected and pre-analyzed data.
Applying these qualities helps AI technology to gain users’ trust. And if they trust a service and accept its help, users are willing to integrate it in their daily work routines to easily get better business results.
AI and UX impact - helps users make informed decisions faster and provides more confidence and oversight.
UX combined with AI: A strong team
To gain full user trust in an AI service, AI/UX design needs to cover not only user interfaces, but also displaying information in a use-case conform manner.
AI/UX must treat algorithm results in a transparent and explainable manner. This allows users to build their trust in the technology over time and accept it as support not as intellectual competitor.
Why AI needs to invest into user trust
To foster this trustful relationship, SAP has published the ethical AI guideline to provide guidance when implementing AI in your business solutions. This policy defines the rules, expectations, and direction for the life cycle of development, deployment, use, and sale of our AI systems.
Why is trust such an important quality in the context of AI? Talking about AI-supported software means to talk about a transition of responsibility – away from human-driven decision processes towards machine-made decisions. This entails human users are handing over their decision-making and control towards machine/AI calculated decisions.
You might argue that such cases were already applied in the past. This is true. And looking back to the history of ERP software, cumbersome manual processes were in the specific focus of digitalization and automation: From data processing via punch cards to ERP software guidance using predefined decision flows with extended rule-definition engines to take one example.
It’s important that AI integration means a new quality of learning and decision-making, with software taking the driver’s seat. AI-enabled software is more independent; it takes its decisions always with the goal in mind to not only deliver but to optimize, alter, and adjust to changing business conditions.
How UX becomes the key to user trust
At this exact point of an AI-controlled environment, UX has an essential role. It guides the way how information is structured, displayed, and made interactive for the user. By introducing users to AI services, UX has the obligation to create this trustful relationship users are looking for. It’s an obligation because UX isn’t only the domain for creating simple and easy-to-use software, but also to care for human needs. Keeping users in focus means the responsibility for UX “to create a truly human-centered AI” as Tim Brown and Barry Katz state in their book “Change by Design”.
Transferring this to a concrete example, SAP SuccessFactors has developed an AI-supported service for a guided training selection based on users’ interests. Users first enter their training requests and interests and get a personalized selection of training course suggestions supported by machine learning in return (see SAP Help for more information).
SuccessFactors training proposals supported by machine learning
Let’s take this recommendation example and transfer it to a recommendation service for logistics focusing on material deliveries. As inspiration serves the unforeseen event of the 2021 blockage of the Suez Canal in Egypt. The effects were heavy logistic delays, for example on material deliveries. Material providers or delivery processes could be adjusted in an automated and fast manner – all enabled by AI avoiding material shortages. In this context, you might continue your read about our Situation Handling example with a focus on efficient material handling. AI support could suggest many faster alternative solutions than human actions.
With this background, beneficial effects of AI-supported decision-making processes become obvious. AI enables businesses to react faster, with solid solution proposals for determined problem situations.
To keep fast and autonomous AI decision-making processes under human supervision with the previously mentioned qualities (control, transparency, and explainability) are easy to understand in this context. Users will gain trust in AI-driven software only if it treats them as humans and shares transparency on its AI-made decisions.
This kind of reasoning about the usefulness of design isn’t new. It actually dates back to the very beginning of the design profession. Already 76 years ago back when computers were invented, László Mohogly-Nagy (teacher at the German Bauhaus and founder of the Chicago Institute of Design) stated: “Design is a complex and demanding task. It entails the integration of technological, social, and economic requirements […] it’s about thinking in relationships.” Fruitful relationships can only grow in a trustful environment. This becomes even more important as we’re creating machine services to take over human decision-making processes.
Additional References:
Acknowledgement
Many thanks to my teammates of SAP’s AI Design Practice for feedback.