Interaction Discovery in Agile UX

Project background


The Interaction Discovery project explores the problem of how our designed user experiences can be made more resilient to surprising interactions, when people use something in unanticipated ways or for unexpected reasons. In recognition of its widespread use, we are focussing on how this can be done in the context of agile development.

Challenges in the Discovery of Unwanted Interactions April 2019

The question of how to discover unwanted features of a design can be broken down into a set of closely related questions (Fig 1).

If user research identifies pain points, how relevant a prediction are they for a new or modified system that does not exist yet? Current use may also include important workarounds developed by the users themselves if elements of the system are programmable in some way, raising questions of customisation. If a problem is suspected, evidence might be sought through comparative testing, but how would the alternatives be chosen when some of them are undesirable? Would they be sufficiently different for the suspected effect to be seen in typical use, or would an unusual scenario be needed for differentiation? Is it practical to write a clear test hypothesis when exploring a problem, or would it rely too much on luck and imagination? When considering the resilience of a system, recognition of how you are currently succeeding can be important. Use of modelling and simulation can be useful, but raises questions of how reliable indicators will emerge from those models, and whether representation of the user population is adequate at the start and how its evolution should be managed to track contextual changes. It may be difficult to anticipate the user's navigation of a future system, and any extrapolation of current user stories to identify future issues may have validation problems.

This breakdown has been used as a guide in designing structured interviews to address the question.

A breakdown of possible methods of discovering unwanted interactions
Some options and challenges for unwanted interaction discovery

Interaction Discovery Research Questions (Updated) July 2020

Some industries have products independently assessed before delivery and address shortfalls by imposing limitations on how they are used rather than making late changes. The motivating question that underlies our study arises from participation in these evaluations:

Why do experienced teams of suitably qualified people still deliver products that fail to meet their non-functional engineering and usability goals?

Our study is an exploration of what designers do to identify undesirable interactions. The interaction type of interest is directive control, where users issue instructions to the system. The more complex situations that arise in objective control, where a goal is set but the system has a degree of freedom in how it achieves it, will not be considered. Similarly, conversational interactions, as used in chat-bots, and the manipulatory or exploratory interactions used in virtual environments will not be considered.

The idea of interaction discovery as a mindful examination of cooperation assumptions and requirements between users and systems will be developed and informed by the research.

These ideation activities come under additional pressure in development environments based on the rapid introduction of functional increments, making Agile practices of particular interest as both a challenging technical problem and one sufficiently bounded in time to facilitate study. Our primary research questions address the particular challenges of interaction discovery in Agile development:

What methods are applied in current Agile software design practice to discover interactions with the user that the intended users will consider undesirable?

Users may have concerns about the integrity and use of their data. The related question of interactions that the system operators regard as undesirable might require consideration of commercial, security and availability issues that do not necessarily coincide with these concerns. Where it is necessary to make the distinction, our study will consider only what is desirable from the point of view of the intended user. Interactions that are immediately undesirable at the time of the instruction, and those likely to be undesirable within the timeframe of the overall activity, will both be considered.

One method of encapsulating the behaviour of an identified segment of the user population is to link their behaviour, their needs, and a narrative of use, with a stereotypical user description, or persona. The way the data is presented may affect the choices made, so we have an interest in personas and other ways of understanding user research that designers create for themselves, as a part of an implied choice architecture, and its possible role in interaction discovery:

How can Agile teams be helped to create and maintain a choice architecture for their work that assists prediction of undesirable behaviours

These questions are embedded in the broader issue of how user centred design is currently practiced in general and in Agile development environments in particular, so addressing our specific issue will also contribute to an understanding of current practice.



By a shortfall we mean something that is disappointing compared to the expectation. This could be a functional, or capability, shortfall if it is not possible to do something that should have been provided, or a non-functional shortfall if it is possible but with a lower quality or performance.

directive control

By directive control we mean step-by-step instructions, typically with only one correct interpretation.

objective control

By objective control we mean situations where a goal has been set, possibly with more than one correct interpretation, where it may not be known in advance how the goal will be met, and possibly will not be obvious after the objective has been achieved what the steps taken were or why they were chosen.