Designed by Machines, Inhabited by Humans: Synthetic Empathy and AI Architecture

About the Project

Architectural design is increasingly influenced by artificial intelligence (AI), with generative models such as Midjourney and DALL-E, and parametric BIM tools assisting designers in the creative process. However, while AI can create convincing and aesthetically interesting spaces, it raises the question whether these algorithmic designs can evoke the same emotional, cognitive, and empathetic resonance as those created by human designers.

According to Pallasmaa (2015) architecture, is an empathetic act: it is an attempt to anticipate how others will experience space, light, and material and the design process relies on the designer’s capacity to “project oneself into the space of another.” Yet, the advent of AI-driven creativity challenges the assumption that empathy must originate from human intention. The term synthetic empathy has recently emerged to describe machine-mediated approximations of affective understanding. In architectural terms, this raises critical questions: can AI models, trained on aesthetic datasets, generate designs that feel human? How might we measure such emotional effects scientifically?

This research will address the growing gap between computational creativity and embodied human perception, offering a new framework for understanding empathy, emotion, and authorship in the age of AI-assisted design. It will test whether such synthetic empathy can be detected at the level of human response. Can a spatial composition produced by a neural network elicit the same physiological signals of calm, awe, or belonging as one produced by a human architect? Or does genuine empathy in design require human intentionality?

To explore these questions, six architectural environments will be created. Three will be conceived and modelled by architects using Autodesk Revit, guided by traditional principles of spatial composition, proportion, and material expression. The remaining three will originate from AI text-to-image generation tools (Midjourney and DALL-E), translated into three-dimensional geometry using Blender and imported into Revit for equivalence of scale and navigability. All six environments will then be transferred into a VR engine, allowing participants to explore them immersively. The typologies will be matched to ensure that differences in user response can be attributed to design authorship rather than function or complexity.

Aim

To investigate how human users emotionally and physiologically respond to AI-generated versus human-generated architectural spaces, and to develop a conceptual framework for “synthetic empathy” in architectural design.

Objectives

1. Design and construct six immersive 3D architectural environments using BIM/Revit and VR platforms: three designed by human architects and three generated through AI text-to-image and 3D workflows.
2. Measure and compare participants’ neuro-physiological and self-reported responses to these environments, focusing on indicators of empathy, comfort, and engagement.
3. Identify spatial and visual factors that most strongly correlate with emotional or empathetic resonance in each environment.
4. Develop a methodological framework for the measurement of empathy in architecture, integrating tools from neuroscience, psychology, and computational design.
5. Propose new design principles for embedding empathetic responsiveness into future AI-assisted architectural workflows.

Ultimately, Designed by Machines, Inhabited by Humans seeks to reveal how empathy, whether human or synthetic, is encoded and perceived through architecture. By comparing how people feel within spaces conceived by humans and those generated by algorithms, the study will interrogate the role of empathy as a medium in the dialogue between designer and inhabitant, which may yet be shared, translated, or even reinvented in collaboration with the machines that now design our world.

Supervisors

Dr Athina Moustaka

Professor Apostolos Antonacopoulos