Some initial ideas swirl around the metaphysics of design; using the Wikipedia entry, an examination of the fundamental categories into human understanding. (see metaphysics)
If design is fundamentally about a object, process, or system to serve some kind of purpose (see design), within functional, aesthetic, economic, environmental and socio-political goals-constraints, then the possible processes and assumptions which shape and construct the design process-thing-context milieu are important and numerous.
There has been numerous design approaches for this construction <enumerate a number>. The sheer creativity and number pose both opportunities and challenges for design going forward, especially when considered in the light of recent advancements in artificial intelligence.
The purpose of our paper is to take both a step back and forward by eliciting a number of diverging assumptions about the process and purpose of design, exploring common and competing metaphysical assumptions about the nature of and goals for it.
By metaphysics, we take a Kantian position of exploring a few key fundamental categories of understanding which direct design. These categories are metaphysical in that they cannot be proved with complete empirical knowledge or airtight rational justification. For example, the need to include the users affected by technological design is a metaphysical assumption about what needs to happen (for success), since it posits “users” as a important category, with designers and managers as implied and important other categories. It thus imposes a view of the world onto complicated situations by directing people's attention to particular categories, and with it, particular determinations and actions.
The “users” is also metaphysical in that the empirical justification and rationale for categorizing people in this way and facilitating participation from them is always incomplete. The ebb and flow of its nature, purpose and usefulness to varying design processes and goals, across time is evidence of its metaphysical. For example, users and participation were considered necessary to address unionized workers (and resistance) to large production systems in the 1980s; considered necessary for organizational success through increases in quality-of-working life during the same period through socio-technical design; was necessary to ensure technology acceptance of ERP systems in the 1990s; was a problem throughout the same period because of the perceived expansion and chaos of technology resulting from end-user computing and bring-you-own-devices later on; is no longer useful because of the availability of numerous software apps that can be downloaded and utilized by users themselves.
From just this quick list of approaches to users and participation, there are a diverging set of core metaphysical assumptions about the nature, process and purpose of human participation in technological design.
Our aim in this paper is not to highlight and resolve these tensions, but to illustrate how highlighting and holding these tensions at the same time are helpful and necessary to address complicated design situations. This approach is informed by key ideas from dialectics, postmodern approaches to contradictions, and the possibility for wider vantage points through reflective judgments from experiences across participants.
As examples of this approach for participation: how could the playfulness of the software app approach be possible in the construction of organization-wide systems? Where and how is the development of technology and increased quality-of-working life in harmony and conflict with organizational efficiency? How can systems use be discerned and designed early in design through simulations? How would such simulations assist both the familiarization with and design of emerging technologies? How would people as “work-designers” or “experience-designers” change how we approach participation in design?
These and other approaches have been explored in various ways elsewhere <cite>. Our contribution is in eliciting and juxtaposing a number of key and contradictory metaphysical assumptions which could lead to a richer and human context for design.
In this paper, we illustrate this cross-metaphysical thinking of participation in design by examining important metaphorical assumptions in the design and participation of AI systems. We explore this in detail through the development and use of AI systems to fund at-risk students. Some key points:
the design of the AI system involved two technical design processes that can be informed and misinformed by traditional and singular assumptions about participation, design, and implementation. In this case and in many other AI cases, the projects begin with the selection and use of relatively well-established mathematical methods by experts, developed over many decades, now increasingly used in data analytics. The choice of these algorithms remains largely a specialized endeavour, although attempts to explain and engage a wider public to understand it is on-going. There is also a second algorithm produced by the first, through the patterns in large data sets, typically through statistical weights. These weights are shaped by the first algorithm through supervised (humans dictating the parameters) or unsupervised (letting the algorithm produce patterns without human imposed labels) learning . If the secondary patterns are “supervised”,
then people (experts and possibly government oversights groups and school administrators) were involved in determining how the first algorithm would be guided in its determination of the patterns through the selection of parameters; if unsupervised,
then the patterns emerge under little human influence and guidance. The algorithm develops whatever patterns in the data through various understandable and less-understandable combinations that are strongly correlated. The main site of influence in this unsupervised case would be the data and its initial structure, and the mathematical algorithm, which despites its semi-independent state, is a human construct. In both the supervised and unsupervised cases, the role and nature of humans, machines and decision making has shifted and stretched across time and space, perhaps radically, captured by but also escaping typical metaphysical assumptions about each of them, especially with the term “machine learning”.
Can machines learn? If so, how does this learning compare to and differ from human learning? Have humans ever learned without the aid of other devices? Perhaps at this point, we can ask the machine to give us an answer…. Human influence (as another position of participation) not only differs in how it shapes the first and second algorithms, but also the before, during and after machine learning and it's use. In terms of before, the data collected and used by the algorithm, of many students and from many states, is now an important influence on the patterns. It is these resulting patterns from the data that matter, and less the technology or the software algorithm as was previously the case in the design past. It is this other humans and their resulting data, from students and administrator, generated by them using other technology, that influences the data “before” AI design. Surrounding the before and after is a “during” phase which involves the collection, storage and flow of information through other human-influenced social-technical assemblages that facilitate, change and restrict the flow of data. For example, privacy laws can restrict whether data flows at all, and whether people have a say in opting in or out of their data and how it is prepared, summarized or detailed, and used. Once released, other people are involved in moving and “cleaning” the data for potential use by the first mathematical algorithm, through various human directed tasks to render it possible to use by the mathematical techniques in the first algorithm, to generate patterns which will become the second algorithm. <more> And finally, there is the “after” phase, where the secondary patterns are used in particular situations and cases, producing new data and predictions. These data and predictions are then used by other people to render important judgments and decisions about other people and situations, which then affect and influence other people's lives. The results of these interactions may eventually become days for the next AI design process.
And so for example,
Within and across these approaches,
If we took that to design and participation, then we could argue both the necessity of and “excess” (to use Derrida's term) for some key and fundamental ideas in design.
See Coyne, R. (1995). Designing Information Technology in the Postmodern Age: From Method to Metaphor. The MIT Press. https://doi.org/10.7551/mitpress/2373.001.0001, chapter 3, Deconstruction and Information Technology
See some important thoughts on metaphysics, deconstruction and Derrida at Coyne_1995