Astrocytes and blood vessels within the retina (Photo by Sergi Reboredo/VW Pics/Universal Images Group via Getty Images)
Artillery Row

Architecture’s abysmal ignorance

Artificial Intelligence reveals what experts deny

Recent scientific applications of Artificial Intelligence (AI) amount to a paradigm shift in settling the architectural debate. For decades, architects have been arguing against common sensibilities about styles, preferring — and fanatically insisting upon — anxiety-inducing structures that almost everyone else hates to experience in person. The criteria offered are that buildings and surfaces have to be judged by how “cute”, “innovative or “pure” they appear in a model or rendering. Forget how you feel close to them and when you have to actually approach, enter and work inside. 

This dark era of treating people as social insects may be ending. Unexpectedly, technological tools originating in the commercial sector may turn into the means of liberating beauty and emotions from decades of suppression.

Objective beauty is biologically-based 

Both Christopher Alexander and David Deutsch discuss objective beauty in terms of universal patterns. Revolutionary developments in understanding the human brain bring us to an appreciation of objective beauty that is rooted in the mathematics of complex organised forms. Those special configurations arguably gave rise to the first life forms, which then evolved into more sophisticated plants and animals. Our DNA encodes complex visual templates to which we respond unconsciously. Subjective beauty, on the other hand, can be anything that people learn, but it is an acquired or imposed preference.

The common person, equipped with a superb neural system as the result of millions of years of animal evolution, cringes at large structures (i.e. buildings) that communicate a psychological threat. There are sound medical reasons for not ignoring such a negative reaction, since experiencing long-term stress is known to generate pathologies in the body. 

The hegemony of the architecture-industrial complex — consisting of building financing, global construction, media-certified “star” architects, real-estate speculators and a machinery of architectural education resolutely stuck in dogma and ideology — crushes public opinion. It has managed to silence the occasional protesting voices for decades. 

We finally possess powerful tools to finish off inhuman architecture

Artificial Intelligence (AI) is now capable of judging what sort of buildings and environments are good for our health and psychological wellbeing. The tests are ridiculously easy to perform. We finally possess a set of powerful tools that are ready to finish off a century of absurdly inhuman architecture. 

Distinct AI applications have immense value for the future of architecture and humankind. These recent developments are mostly unknown to the public at large. They were anticipated by architects and theorists marginalised by dominant architectural culture, but for the régime, they are certainly unexpected even as they pose an existential threat. 

AI performs a preference survey involving a billion people 

Text-to-image software was trained by AI on an enormous number of images of buildings, so that verbal prompts such as “ugly building” and “beautiful building” generate a novel photorealistic image each time. Any reader can perform virtual surveys at the click of a key. One has a choice among competing and distinct software packages, including Craiyon, DALL·E 2, DreamStudio and Stable Diffusion. The results all agree, settling decades-old debates since “ugly building” produces industrial-modernist high-rise buildings low in objective beauty, whereas asking for “beautiful building” produces traditional-looking buildings high in objective beauty. 

A second, related type of big-data survey uses AI-trained software to produce realistic paintings of an “ugly building” versus a “beautiful building”. This technology advances AI into the creative art world of representational paintings, as exemplified by the program Midjourney. The same two prompts generate drab and structurally top-heavy brutalist buildings low in objective beauty, versus colourful and curved traditional buildings closer to Art-Nouveau high in objective beauty. 

The sheer number of data that AI uses to create those images represents the preferences of a sizable portion of the world’s population, which overwhelms the dominant but extremely narrow authority of so-called experts dictating architectural style. The fact that at least five independent surveys give consistent results adds an incredible degree of verification to these methods. 

Diagnostics from eye tracking and visual attention scans 

Eye tracking used to involve bulky and heavy instruments fixed in a laboratory decades ago, but it is now miniaturised into a set of eyeglasses. These can measure where the eye spends its time in scanning an image for either a real, physical setting or a photo shown on a computer monitor. How we react to a building or urban setting is determined unconsciously by this quasi-instantaneous information. Measuring the first, pre-attentive fixations indicates a person’s reactions that instinctively decide among the three automatic responses of fight, flight or freeze. 

AI goes one step further in training software to mimic actual eye-tracking responses. 3M Corporation’s Visual Attention Software (VAS) simulates eye tracking to within 92 per cent accuracy. The software gives an instant “heat map” showing where the eye is predicted to gaze on an image. This program was developed for commercial applications such as advertising, and it has been applied to evaluate the visceral engagement with architecture in a series of studies: the attractiveness of building façades in different architectural styles, and what entices people to walk in a particular direction. These results confirm the AI-generated visual preferences described above.

Visual attention comparison scans reinforce an independent 2020 Harris Poll that revealed overwhelming public preference for traditional versus modernist US federal buildings. A vicious and partisan media debate accompanying the announcement obscured the result, despite it showing that preference is independent of political party affiliation. Afterwards, two separate groups of researchers verified the results of the Harris Poll by using direct eye-tracking experiments and eye-tracking simulation software

A pathetic reaction from the architecture-industrial complex 

Whilst mainstream architecture talks about AI technology, it addresses none of the above findings. Instead, those programs are seized as possibilities to continue the architectural establishment’s control of world architecture. Just like any power system, extractive global construction is interested only in perpetuating its profits, in this case by continuing to impose architecture top-down. 

We have to rely upon AI to recover senses that our bodies lost

AI is being misapplied to generate ever more monstrous designs. No thought is given to how such structures will affect the health and wellbeing of users; it is assumed that any person who objects is an ignoramus. Even august professional societies have fallen to using bullying and propaganda as a way to silence critics of the establishment. 

The problem of architecture is no longer about styles, with their toxic ideological and political associations. It concerns public health and psychological wellbeing. Vested interests and sclerotic organisations regulating antiquated architectural education need to change. New scientific tools discredit those who seek to maintain their comfortable stranglehold on public taste by ignoring human biology. 

Monumental errors that could easily be prevented

The public has before it a group of decisions on new buildings, along with additions and renovations to older buildings. Unfortunately, the old stylistic arguments often lead to the wrong choice that is regretted for generations afterwards. New diagnostic tools can help interested parties decide in a few minutes. Approving an addition to a national monument that is widely considered to be insensitive (or even vandalistic) should be based on visual diagnostics. It does no good to refer to the presumed qualities of the architect, especially to what he/she declares in self-serving promotion. 

Visual attention studies immediately expose design mistakes. New industrial cities that were hyped as technology hubs fail as human environments because users could perceive them as menacing and ugly. The human experience is disengaging. Out of many such blunders — the result of dominant architectural ideology rejecting historical lessons and traditional precedents — we cite Boston’s Seaport technology hub and Shanghai’s Lingang industrial district

Criteria for which buildings to preserve, and which to demolish and replace, switch totally. We dump the historical argument, consistently misapplied to anoint brutalist horrors simply because they are old. Any building that generates visual stress is harming us. Pulling it down is not a return to a fairytale pre-modern past, but the necessary step towards creating a healthier contemporary city based on scientific knowledge. 

Symmetry and objective beauty in faces 

Yet another related AI tool has been developed for evaluating facial beauty. Software based on Google’s NIMA program (Neural Image Assessment) is trained on an enormous dataset of what is generally considered beautiful. Despite concerns with bias and stereotyping, the cosmetics and facial beauty industry is applying AI-based personal beauty measures. People unconsciously display — and seek in others — biological features that signal health. Throughout history, and despite wide variations in tastes at any particular place or time, enhancing natural beauty to express health was sought in improving one’s attractiveness. 

Human perception seeks complex face-like symmetries in buildings. Not finding any hinders emotional attachment. The unconscious reaction to facial beauty — beginning from a baby’s recognition of its mother’s face — is biologically akin to that from an architectural façade. Architects trained in the modernist idiom reject human evolution that imprinted nested symmetries about the vertical axis in our innate visual preferences.

Emotions distinguish objective beauty from ugliness 

There is more to the technological tools discussed thus far, which perform as unbiased and extremely powerful assessors of visual information. Visual attention indicates engagement, which combines with the emotional state of the subject whilst experiencing architecture. The emotional experience of connecting visually to a place can be measured using portable, wearable indicators of mental health, mood and stress levels

The joy of the architectural experience is emotional and immersive: a kind of mystical union. A beautiful structure pulls your heart as does a beautiful person. This is what generations of self-appointed “experts” suppressed through clever manipulation and propaganda. Ironically, we now have to rely upon AI to recover senses that our bodies lost through deliberate numbing. 

Portable sensors detect subtle changes in facial features, which can be linked to induced stress in the body. Ugliness — as visual information that triggers alarm in the neural system — creates stress, which distorts baseline facial expressions. This mechanism permits architectural “ugliness” to be detected indirectly. 

New headsets developed for use in the metaverse now incorporate such facial feature detectors as standard components. Companies offering diagnostic products for facial expression analysis include Affectiva, iMotions and Noldus. These tools sense how detailed facial features respond to an image, supplement that information with other biometric sensors, and use AI to infer the subject’s unconscious emotional responses. 

AI company Emotiva used webcams to measure facial muscle movements in developing the “Stendhal Index” for evaluating artworks objectively. Reasoned Art Gallery selects digital artworks according to their beauty rating on the Stendhal Index, which it then exhibits and sells to the public online. A separate application, Naiada software by Kellify company, classifies fine artworks according to their AI-perceived beauty. AI-generated artworks are traded using Non-Fungible Tokens (NFTs) in the virtual marketplace of Crypto Art

Existing applications are in the realm of fine art, and the same methods will rank objective architectural beauty next. Nevertheless, century-old fraudulent arguments continue to deter people from demanding an emotionally-healing environment. They choose to ignore science and remain subservient to thought control. Architectural fashionistas fool the gullible public with more glitz and marketing, manipulation via emotional triggers, hypnotic promises of “progress” and the implanted terror of “regression”.

Enjoying The Critic online? It's even better in print

Try five issues of Britain’s newest magazine for £10

Critic magazine cover