Building on the work I did at Venture III, I designed and delivered a 2 days workshop at IBM in London. Working with participants from the UX and data science teams we set out to think about how design could be used as a language to bridge the two disciplines.
The recent development of machine learning as a decisive technology in the field of digital products and experiences has depended on increased computing power and access to massive data sets. While not new in itself, machine learning is set to have a profound impact on how people experience the world through computational systems. Due to this expansion in the availability of training data and the increase in processing power, machine learning (ML) algorithms now determine much of what we see online, whose opinions we are exposed to, the cost of the goods we buy, and the hospital treatment we receive. The discipline of user experience design (UXD) has faced a number of challenges when responding to this phenomenon. Among these are explaining the complexity, uncertainty and opacity of ML technologies to designers and devising ways for UX designers and ML data scientists to effectively collaborate on new designs.
The call from HCI researchers and design researchers in this context (Yang et al. 2018) has been for ‘sensitising concepts’ that will help make ML available to UX practitioners as a new design material. With roots in social theory, the idea of sensitising concepts has been adopted in design to help open up space for future design innovation. Sensitising concepts are embodied in design artefacts that reach beyond their immediate material manifestation to sensitise other designers to the possibilities of the material they are made of. Sensitising concepts expand the field of practice of a particular design domain by demonstrating how new materials encountered within that domain may be used. Sensitising concepts are embodied in ‘designerly abstractions’ which free designers from having to fully grasp the technical constraints of ML technologies, instead allowing them to explore alternative forms. They act as boundary objects between UX design and ML data science fostering new ideas and bringing the gap between design possibility and technical capability (Yang et al. 2018) working to make ML available as a design resource.
Working in pairs, participants started by drawing visual narratives of how they perceived ML in digital products. The intention was to ask the question: What is going on? i.e., What work does machine learning perform in digital systems? These were done from the perspective of users, rather than designers or developers. The topics chosen were driven by some of the work done by participants i.e. on conversational AI systems and automated banking. This initial phase is intended as an introduction to the workshop and we therefore to focus our analysis on the later phases.
The second phase of the workshop involved participants distilling the stories from phase one into distinct characteristics that can be expressed in a superhero costume. Participants use wooden human figures to express positive and negative attributes of what ML does in digital products. Physical materials include coloured acetate, felt strips, paper, and wire. The aim of this phase was to get ideas off the page and into tangible, substantial forms around which shared concerns could emerge. Initial ideas were informed by the visual narrative outcomes and by the positive-negative prompt.
The third phase of the workshop required participants to work together in pairs to represent the scale and complexity of ML systems, thus moving on from effects of ML to direct representation. The materials available for this phase were unconstrained. The resulting models built on phases one and two by examining two specific aspects of ML that are often seen as difficult to grasp and communicate, scale and complexity.
The final phase of the workshop required participants to imagine speculative digital products in which their observations and thoughts about machine learning could be implemented. This phase was a chance to consider how the abstract thinking in the previous phase could be implemented in a real system. The aim was to explore how the evolving shared understanding of ML could be materialised in physical forms and then communicated and discussed. The focus for this phase was on product development, bringing together the previous steps to consider how the company might respond fifty years in the future to the opportunities of ML technology.
One key factor emerging from the workshop was that company structure and organisation often fails to support communication and collaboration between departments. Data science is seen as inhabiting a completely different conceptual and professional territory to design. Practitioners in both departments rarely meet and are expected to undertake different tasks. This is evident in two ways revealed by physical models. Super Data Guy, a costumed wooden figurine is unable to fly. The figure is enclosed by two layers of constricting material and is weighed down by heavy feet. Like Superman the figure indicates a desire to take flight by raising one arm. Both data scientists and UX designers agreed that the systems that would allow them to freely imagine the possibilities for ML technologies do not exist. There is no space; either physical, conceptual, or organisational for exploratory imaginative activity to take place. The workshop was thus the first opportunity for participants simply to meet and talk let alone engage in an interdisciplinary collaboration around ML technologies. This reveals that there is some important work to be done at an organisational level in supporting and scaffolding collaboration between data science and UX departments before meaningful interdisciplinary inquiry into ML can be done.
Micro/Macro shows how the distribution of computational and organisational resources both reflects and determines the values they are assigned in a system. Considering data science knowledge as a system (i.e. company) resource that accrues high value over time means that it ends up prioritised over design resource. Participants reflected that resources arranged in a hierarchical manner optimises commercial operations while devaluing interdisciplinary collaboration.
Future work in this area will involve working with different types of organisations. Situations where data scientists and UX designers work in the same company and building are still relatively rare. I expect this to change quickly and for the concerns explored in this workshop to become more widespread as ML technologies spread into other areas of human activity. Access to the means of production of these technologies is important for designers but also for other disciplines such as strategy, planning, and policy. Future work could also include more participants from adjacent and contrasting domains in order to enrich the discourse around ML and society.