It is our role, as designers, to ensure that users can effectively interact with technology in ways that give them a sense of control and agency. However, when a user stops being the sole operator and becomes a cooperator with technology, they also grow vulnerable to the system's actions and limitations.
That's why designing agentive systems is a delicate balancing act, mainly because it requires that both users and intelligent agents interact strategically with each other to fully realize their capabilities and goals. This piece will describe some of the design and research methods I've applied to approach the tension between users and technology.
As a UX consultant, I collaborated with various startups to conduct user research, build prototypes and validate MVPs. By working closely with key stakeholders, I translated their knowledge into insights and delivered the correct information to those who needed it to make product decisions.
Because designing for agentive technology has its unique challenges, designers must adapt the methods we choose to the user and technology's nature and the particular problem at hand. Choosing a proper methodology requires understanding the potential factors influencing the user experience and recognizing the knowledge gaps that will prevent us from making informed decisions.
Methods, when adequate, should alleviate the main stumbling blocks that prevent us from delivering optimal design solutions:
Often, domains have their own paradigms, details, and considerations which can influence the design, development, and testing of solutions. Not knowing enough about the problem domain to argue for particular approaches can hinder our ability to make good decisions.
Complex problems will most likely involve users with particular requirements. Thus, designing the right solution for them will depend a great deal on our understanding of their needs and our ability to describe them.
Although problems are best defined in an abstract sense, some are so dependent on their implementation context that they require a more profound understanding of the technology itself.
Complex problems often exist in uncharted territory, where learning resources and references are often unavailable or insufficient. As a result, designers often face a broad range of uncertainty about the technology's impact on the users and their goals.
Assessing these factors in the early stages of a project can help establish the proper methodology and increase outcomes predictability.
Agentive design is the design of technology that assists humans in achieving their goals. Contrary to popular belief, the design of these systems is not exclusive to robotics. It can be applied in a wide range of fields, for example, in medicine, legal contexts, banking, and education. Consider, for example, the possibilities for applying agentive technology in a business context where operations have a dynamic, multi-layered nature. Such systems could examine and diagnose problems as they arise, use assessment tools to optimize their responsiveness, automatically generate decision support resources, and more.
In my experience designing agentive technologies, it is much more than simply using technology to help people. It's about making it so that people can get on with their daily work.
The following is a high-level outline of a methodology for designing agentive technology products and services. It covers some of the initial design research activities, the definition of the end-to-end experience, and the solution's actual testing in the real world.
A literature review might sound like an unlikely start for a design project, but in fact, it is a powerful way to start digging into a problem. Research papers, textbooks, and articles are beneficial for contextually locating relevant (and often overlooked) data and insights. In the case of agentive technology, this could involve analyzing the workflow that will be impacted by the product, the pitfalls, problems that the current solutions miss, etc.
Knowing what makes users behave in a certain way or determine a particular outcome can be a challenging endeavor. One way to do this is by pinning down what makes users do something in the first place. That is, by probing into the moment-to-moment, on-the-spot nature of their behavior. Contextual inquiry, or users' observation in their environment, can be an invaluable starting point for developing agentive technology products.
A process map is a tool that visually shows relationships between several inputs and actions taken in an environment. It can help gain insights into how the users will interface with the agent and its features from a design perspective. Similar to contextual inquiry, it offers an excellent starting point for exploring user processes in complex systems.
Rapid sketches or wireframes are usually a good way to start exploring a design problem. Still, they can be limited in how many scenarios can be simulated, and in the case of agentive design, these can be numerous. To ensure that problems are thoughtfully explored, an approach based on generated scenarios is often helpful. Suppose we know the universe of possible user inputs and the controllable characteristics of the agents. In that case, we could generate a realistic set of situations and use them to explore design issues.
Tools like spreadsheets, or similar software, can be used for the automatic generation of scenarios. Likewise, interactive whiteboards, such as Mural, can be used to document scenarios and connect them to others.
I collaborate with my clients in the research, design, and development of products and services that use AI to help their customers achieve their goals. Reach out to discuss your ideas and how we might work together.
Ghislaine Guerin 2021