The Architect is Dead: Long Live Architecture! No, false alarm: There can be no architecture without architects, no matter how far AI is advancing. The more AI tools for architectural design will be developed, the more architects will be empowered to concentrate on the sophisticated, creative and reflective aspects of their responsibility and the more they will be able to expand their ambitions, i.e. their speculative-predictive reach and agency with respect to the built environment’s social functioning. My research project, “Agent-based Parametric Semiology,” contributes to this computational (and methodological) empowerment of our discipline in an aim to advance our discipline’s capacity to simulate the way spatial arrangements shape social interaction processes, i.e. the final purposes of our space-making investments. This envisaged Life Process Modelling is much more ambitious and general than currently available crowd modelling systems in that it goes far beyond circulation problems to encompass the full richness of user interactions.
Since all of human action takes place in designed spaces, surrounded and empowered by designed artefacts, the scope of architecture and the design disciplines is comprehensive, and so is, by implication, the project of Agent-based Parametric Semiology. The design disciplines claim universal and exclusive competency with respect to the global built environment and the total world of artefacts, insofar that these environments and artefacts function as interfaces of communication. This definition of architecture/design’s societal function – which I have elsewhere condensed into the formula ‘design frames communication’[1] – excludes all merely technical (under the hood) aspects of the environment’s functioning. This emphasis on communication and its distillation as our core competency/specialism as architects/designers is congenial to the character of our society as knowledge, information, or network-society. All work, indeed all action, is communication. All spaces and all artefacts are communications too. As architects/designers we are concerned with the informational content and communicative capacity of spaces and artefacts. This explains the prominence of semiology within my conception of architecture’s task. According to my theory architecture and the design disciplines including urban, interior, furniture, fashion, product and graphic/web design constitutes a single, unified, sui generis societal discourse practice with universal and exclusive competency concerning the physiognomy of the built environment and world of artefacts conceived as communicative frames. All social events depend on our frames: a large sports event (architecture) as much as a dinner date (interior and furniture design), a magazine article (via graphic design), as well as all Facebook exchanges (via web design).
In our agent-based life process simulations we equip agents with a whole stack of behavioural capacities or choices, not only walk or stand. These behaviours are made location or frame dependent, i.e. which behavioural choices are available at any moment depends on where the agent is currently located and which perceptions or informational clues are available. Agents encounter and interact with each other in accordance with the designed communicative spatial framings. An overall pattern emerges, differentiated and ordered through the differentiated matrix of differentially articulated spaces.
Each agent is equipped with a “brain”, i.e. a set of decision rules (decision trees) that link perceptions to actions with various degrees of probability (fuzziness) that might change due to learning experiences. Our agent populations are differentiated by type, or even individualized. The methodology foresees that – depending on the type of institution to be accommodated – prior social relationship valences (e.g. communication history as drawn from email records in the case of a corporate space design) can be taken into account and modelled within the agent population.
Once the described simulation capacity has been achieved the design is varied and varied again while the resultant differential steering of the interaction processes is observed, measured and evaluated with respect to the success criteria that ultimately (but non-trivially) flow from the client brief. The aim is to optimize the spatial arrangement and semiological articulation of activity zones and furnishings in ways that maximally facilitate the relevant encounters, cooperative gatherings and productive interaction scenarios that sustain the institutional processes of the client. Therefore this methodology is finally homing in on the very purpose of building in the first place: to gather people into a productive cooperative process (in the widest sense).
This methodology of varying the design with a user simulation in the loop lends itself to the incorporation of adaptive capacities, including real time responsiveness: sliding walls, mobile partitions, roaming furniture, changing lighting conditions. The easy availability of all sorts of sensors and actuators suggests that the time is ripe for responsive artefacts and adaptive environments. It should be evident that an adaptive capacity could be strongly productivity enhancing in an era where the brief’s (programme) requirements should take the form of “parametrically variable event scenarios” rather than simple schedules of accommodation listing functional stereotypes. Accordingly the design research programme of “Responsive Environments” has been part and parcel of my “Semiological Project” all along. Spatial transformations, scheduled or responsive, are even easier to read as signals or communications than static configurations, as invitation to join a specific interaction situation.
The attendant agent-based life process modelling suggests further that those adaptive walls, partitions, furnishings and lighting systems might be conceived and modelled as agents in parallel to the simulated human agents, according to the same schema of perception/decision/action suggested by the game modelling tools we use: Miarmy, Unreal, Unity. Agents can here also be equipped with changing time – or experience-based internal states that impact on behavioural predispositions. We are led to robotic environments where robots are modelled in parallel to human agents, all equipped with “brains” and all acting in response to internal states, the static environment, other human agents as well as to other robots; here “architectural agents”. Bruno Latour’s Actor Network Theory (ANT) comes to mind according to which human and artefactual actants together form the social system.[2] Foucault’s theory of discursive formations also points towards this.[3]
From here it is only a very small step to imagine intelligent, and indeed spontaneous or creative architectural agents as building blocks of an overall design project (eventually populating our lifeworld). Partitions and furnishings as observant, sensitive and spontaneous guides and animators that actively steer the interaction process, bring people together and move the social scenarios forward. Such artefacts would ideally have the capacity to learn on the basis of experience, real-time positive or negative feedback, or on the basis of performance reviews administered at regular intervals. We can imagine unique personalities to evolve. I think we can presume that for many the envisaged learning capacity, self-directedness, spontaneity would be a positive asset that clients and users would cherish as much as they cherish such characteristics in their human staff. We must not imagine all robots of the future to behave robotically.
The paradigmatic scenarios I have in mind here are large, differentiated corporate office spaces, conference centres, a university campus etc. where many people come together to cooperate and exchange knowledge and skills, i.e. scenarios where human interaction and learning dominates as primary productive purpose and where artificial agency is facilitative with respect to human exchange and interaction, rather than scenarios where sparse human occupation facilitates robotic production as in warehouses and factories. While in the latter cases engineering concerns dominate, in the former scenarios architecture plays the leading role. Here all design is communication design and both environment and robots are designed in view of human perception and comprehension. Although machine vision requirements will constrain the visual articulation choices of the design, this does not call into question the primacy of human-focussed phenomenological and semiological articulation. The subsidiary concern with machine vision plays no different role from any other technical requirement that constrains the search space within which design solutions according to the specifically architectural design criteria, namely communicative legibility, must be found.
In the scenarios I am envisaging as an architect, with robots as architectural agents conceived and designed by architects, these robots/architectural agents remain essentially interface or framing agents with the same immediate social function as conventional, static architectural “agents” like walls, doors, partitions, furnishings etc. designed and evaluated according to the same criteria. My general formula for architecture/design’s societal functions applies here as with respect to all tasks of the discipline: The societal function of architecture is the framing of social interaction, i.e. the innovative ordering of social communication processes via a system of spatial frames. Of course we can imagine further scenarios where we communicate with and inhabit spaces together with robots who do other types of work beyond the work of architectural framing. Earlier I mentioned warehouses and factories… But here human occupation is so sparse and workers are so familiar with their environment that, while never altogether absent if people are at all present, architectural articulation has a rather low priority, like in machine rooms, where perhaps only the control panels are designed as interfaces of communication with a view towards human perception and comprehension. However, we can also imagine scenarios where non-architectural robots inhabit spaces with humans, e.g. in hospitals where robotic diagnostic machines and robotic carers roam around, or e.g. in offices like law firms where training robots or legal expert systems roam around to make themselves available to answer questions. Here designers (our discipline) would still be called upon to design the visual physiognomy and visual or tactile interface with human users. However, the more sensitive, intelligent, self-directed and spontaneous – i.e. human-like – the robotic agents that inhabit and roam our life-world are, the more we might have to consider designing for them as much as we are designing for human users. If embodied, roaming AI-agents inhabit our designed spaces, we must design for them as much as for human users, even if the final aim of all our creative and productive efforts remains the comfort, safety, satisfaction and ultimately flourishing of us human beings. If freedom, prosperity and human flourishing remains the ultimate endgame, each architectural project might have to concern itself with the life conditions, or rather productivity conditions of our artificial but no less intelligent co-workers and companions as a means to this end. Architects would have to take their different subjectivity into account as much as they now take into account the different subjectivities of their various human user groups and audiences. We can treat this as a practical heuristics that does not challenge the “ultimate” ends. However, this makes no material difference to our practice. Our practice – to be effective – will have to treat artificial agents with as much concern as human agents, facing very similar challenges: how to physically constrain and channel the robots self-directed movement (the organisational project), how to take account of the robots’ various perceptual/cognitive capacities (the phenomenological project), and how to inform/instruct the robots about the social character of the environment they are entering (the semiological project).
Although this architectural design concern with embodied AI-agents does not itself contain any ethical dimension, the theoretical reflections this scenario evokes are rather similar and parallel to the ethical questions that arise with respect to the robotic AI-agents. The more we conceive and design artificial agents as learning agents with autonomous decision making capacity, spontaneity and creativity, i.e. the more human-like or person-like they become, the more ethical or moral issues come to the fore. Do such evolved creatures, perhaps with unique accumulated learning experiences and thus “personalities” deserve our respect, the same or similar respect we feel we owe to other human agents? On what would this respect hinge? On intelligence? On consciousness? On the capacity to suffer? No, I think it would hinge on the complexity and uniqueness of the respective creatures’ evolved capacities. This capacity might not be transferable due to the complexity and uniqueness of the robot’s embodiedness as e.g. its capacities might be distributed and dependent on a uniquely assembled cloud network that cannot be copied over to a replica. Replication might be impossible. In this case such a robot (or internet bot) deserves our respect and protection, in the sense that its protection as unique and irreplaceable productive member of our society is in society’s and all our best interest. If irretrievable complexity and uniqueness through evolutionary/historical accrual of features and capacities is the key here, then our historical cities deserve our respect and protection also. They are indeed irreplaceable, inimitable architectural agents we with our intricate social and societal life-processes depend on; and we indeed value, respect and protect them accordingly. The same can also be said to apply to our natural environment. It is now understood to be a uniquely evolved, fragile and irreplaceable creature, rather than an inexhaustible domain from which generic life conditions can be repeatedly forged. Any ethic must – in the final analysis – be pragmatically grounded. The same goes for a professional ethos and methodology like ours as architects. We will be well advised to not only treat our uniquely learned AI environments, co-workers and companions as if they were persons but as architects we must also design for them as if they were equally precious and respected members of our society. The implied distinction between “as if” versus “for real” turns out to be practically irrelevant, and therefore is to be dismissed as metaphysical.