One of my responsibilities at LexX is to ensure that the user experience of our products are best in class. This is so very important for us and our customers because younger engineers entering the workforce have come to expect richer digital experiences. The challenge is that maintenance systems haven’t kept up with the demands of digital natives. This is an opportunity. But the bigger challenge is building user experience for intelligent machines (and software). Design Challenges There are two main challenges we are trying to address. The first is that if we want our users to like our software, we should design it to behave like a likeable person. The expectations from Artificial Intelligence accentuates this need. The attempt is not to de-skill engineers and technicians but for them to feel confident with their decision making rather than making decisions for them. The second, and bigger challenge, is that when systems utilising machine learning (ML) are constantly learning, adapting and renegotiating in a context of evolving autonomous system and human interactions, the design constraints and goals differ from those of conventional UX. Conventional UX and UX for Learning Systems The articulation of the desired future which is at the centre of conventional UX design which is achieved via user research helps designers ideate of divergent framing and solutions to establish what is the right thing to design. With machine learning, the emphasis is what can be accurately learned given the available datasets for a certain application. The shape of a ML mediated future seems to have been mostly driven by the data availability and the learned performance rather than a deliverable vision. In real terms these challenges translate to questions such as what would a journey map look like, would personas be an effective model to understand interactions and needs. How would these interactions work when there are negotiations between multiple autonomous ML/AI systems each with different characteristics. Who is the user, or is the user even an appropriate way to understand the problem. For example, in traditional systems development, developers are used to lengthy user training sessions to train users to work with the systems. In AI systems, the system is the user being trained to work with humans. Behavioural data are fed into the system and algorithms use statistical properties of this data to generate knowledge. An interface then communicates that knowledge to enrich the user’s experience. Finally, interactions during this experience create new behavioural data that can be used to retrain the learning algorithm thus spawning a feedback loop. Emerging Perspectives We have come to the conclusion that there are two fundamental roles that UX plays when designing for ML driven products. These are the interchangeability of problem setting and problem solving, there is no common way to separate problem setting and problem solving in a machine learning process. That is, to purposely use ML to solve the right user centred problem and to design the UX in a way that produces and communications ML outputs appropriately. Matching the machine learning capability and the right UX problem is essential and challenging. So we balance performance and usability to “get the thing right” and “get to the right thing” Secondly machine learning algorithms are designed to look for patterns within a set of sample behaviours to probabilistically estimate the rules underlying these behaviours. This approach comes with a certain degree of imprecision and unpredictable behaviours. Consequently, they require responsible design that considers moments when things start to disappoint, embarrass, annoy, stop working or stop being useful. This stage is, therefore, about finding a user experience that balances between the global predictive power of machine learning and the edge-cases that, in practice, can disassemble the value that users are getting from the product. Our Design Philosophy At LexX we have adopted a design philosophy that balances three things 1) that design is the most immediate way of defining what LexX becomes in the user’s mind; 2) that design is a way we deliver LexX, ie as a service to be experienced of which the product is only a part; 3) recognising ML is the new UX in that UX utilises data to envision and operationalise ML methods solving user centred problems while producing and communicating ML outputs. These principles then lead us to focus on 4 experiences LexX will deliver to every technician in our products every day.

1. Experiences that prepare them for the task ahead. These are experiences focused on creating feelings of assurance by bridging beginning and end of a task.
2. Experiences that foster connections that mitigate uncertainty, improve decision quality through collaboration and build trust
3. Experiences that drives confidence and feelings of empowerment resulting in improved quality of everyday outcomes, decisions and actions
4. Experiences that make technicians feel supported, support that becomes the technicians best ally when under pressure.

If you have a viewpoint on this topic, we would like to hear from you. Written by George Mathew Product Director at LexX Technologies