Koseki-manuscript-happened-20201022
Draft version — Do not cite — Do not share The draft of this paper is based on the layout of Lazer et al. 2009. “Computational Social Science” in Science*
We live in a designed world. The furniture we buy, the bus we take, the phone we call with, the park in which stroll. We plan fields to feed us, and natural reserve to sustain wildlife. We travel the world to see building, and other site which we deemed worthy of preserving. Each of our action builds on cultural signs that can be traced back to previous cultural products and practices that evolved through time and space and defined our lives, organizations and societies.
The capacity to collect and analyze massive amounts of data has transformed such fields as natural and social sciences. But the emergence of a data-driven “artificial design” has been much slower. Leading journals in design, architecture and urban planning show little evidence of this field. But artificial design is occurring–in the GAFAM, and in start ups such as [insert EPFL patent startup]. Artificial design could become the exclusive domain of engineers and computer scientists. Alternatively, there might emerge a privileged set of academic researchers presiding over inaccessible so-called public data from which they build monopolies of moneytizable data processing pipelines. Neither scenarios will serve the human interest of innovation, comfort and beauty.
What value might an artificial design–based on an open environment–offer society, by enhancing the quality of our built and natural environment? What are the obstacles that present the emergence of an artificial design?
To date, computer aided design has relied mainly on small sets of self-defined parametric data. New technologies, such as natural language processing, image recognition and multimodal processing offer a moment-by-moment picture of cultural production over extended periods of time, providing information about both the content and the structure of cultural evolution. For example, color patterns could be designed based on paintings data, and choices about the aesthetic quality of objects could be addresses: should we focus on form or materiality of objects? What shape better fit a given context? What composition produces the most striking visual effect on people? Design integration could be defined over time with generative design pipelines. Such digital infrastructure could define meaningful integration, shapes, colors, materiality, lighting, and other other components of the design project. The data could produce interesting outcomes for objects, buildings, images and interactive plateforms.
We can also define what an overall design strategy for a city would look like, and how to make it evolve over time. Archives have records of atmospherical descriptions from a variety of sources extending over multiple centuries, and paintings and drawings collect picturesque visions of past landscapes. How does this material feeds a comprehensive picture of society-level imaginaries? In what ways can we build on those images to design public spaces and parks? Indoor scenes and portraits of the sane periods thickens the quality of such imaginaries. Allegories and bodily gestures allow portrait past understanding of urban aesthetics. Such data may provide useful epistemological insights: How concepts of good and beautiful evolved through time and space?
Textual archives offer a totally different channel for generating artificial design. Consider, for example, all the material that describe with accuse precision the evolution of historical urban landscapes, as well as the monuments of the past, where the concerns for their virtual presence can become visible for future generations. Oral histories, which by their nature capture a complete record of marginalized groups, offer ample opportunity for design–integrating the experience of minorities into urban built and unbuilt forms. Similarly, audio recordings of music offer a unique opportunity to enrich design project with moods and emotions, whereas Natural language processing offers increasing capacity to define emotional charges and urgency into the design process.
In short, an artificial design that leverages the capacity to create new cultural artefacts with an unprecedented pace and quality is emerging. Substantial barriers, however, might limit progress. Existing ways of conceiving cultural artifacts were developed without access to terabytes of data describing step by step evolution of preceding designs. For example, what does existing architectural design, built mostly on an iterative design process, typically inspired by a few dozen case-studies, tell us about massively longitudinal data sets of millions of building, including materiality, lighting and structure? These vast, emerging data sets on cultural artifacts surely offer qualitatively new perspectives on urban space, but our current paradigms may not be receptive.
There are also enormous institutional obstacles to advancing an artificial design. In terms of approach, two-dimensional and three-dimensional representations present different challenges to conceive and produce. Design must provide them with an enjoyable experience or at least resonate with people. As for infrastructure, the leap from manmade design to an artificial design is larger than from statistics to artificial intelligence largely due to our lack of understanding of the so-called “design process” to start with. There are fewer resources available in the design schools, and even the physical (and administrative) distance between those and engineering or computer science departments tends to be greater than for the other sciences.
Perhaps the thorniest challenge exist on the data site, with respect to representation and inclusion. Much of the cultural artefacts to date have been produce by males of culturally groups. The same goes with the artefacts valorized and thus preserved. To date, digitized cultural artefacts conceived by women, cultural minorities and with differently able bodies remain marginal. To avoid the perpetuation of inequalities, robust monitoring of diversity and inclusion in digital cultural data is needed to safeguard accessibility and usability by all. Evidently, this also concerns all people involved in creating an artificial design. As recent debates over the representation and inclusion of artificial intelligence highlights, the production of inequalities tend to add up in complex data processing pipeline without acute scrutiny and structural diversity.
Because shaping exclusion and inequalities into design can have large scale and lasting detrimental effects on large share of the population, a transdisciplinary enforced sets of procedures, technologies, and rules is needed that reduces this risk but preserves design innovation. As a cornerstone of such a transdisciplinary initiative, design schools must increase their technical knowledge to understand the potential for algorithmic bias and misrepresentation fo because new possibilities do not fit their current paradigm for inclusion. Many design schools would be poorly equipped to evaluate the possibly that complete data could generate detrimental design. Further, it may be necessary for design schools to oversee the creation of a secure, centralized data infrastructure. Currently, existing digitized cultural artefacts data sets are scattered among many instituions with uneven skills and understanding and widely varying quality. Designers themselves mush develop technology that account for diversity and inclusion while preserving data essential for research. These systems, in turn, may prove useful for industry in assessing diversity and inclusion of public space.
Finally, the emergence of a artificial design share with other nascent interdisciplinary fields (e.g., digital humanities) the need to develop a paradigm for digital empowerment of designers and scholars. Design schools need to understand and reward the effort of acquiring high level digital skills. Initially, artificial design need to be the work of team of designers, social and computer scientists. In the long run, the question will be whether academic should nurture artificial designers, or teams of computationally literate designers and design literate social and computer scientists. The emergence of a artificial intelligence offers a powerful model for the development of an artificial design. AI has involved fields ranging from social science to the arts to computer science. It has attracted the investment of substantial ressources to create a common field, and created enormous resources governing algorithms bias in the last decade. We would argue that an artificial design has a similar potential, and is worthy of similar investment.