Embodied Interaction emphasizes that our understanding of the world is grounded in our bodily experiences. According to Dourish (2001), physical presence and sensory engagement are central to how we interpret our environment—an understanding that contrasts sharply with the binary data processing of computers. Human perception is deeply influenced by cultural and personal experiences, which renders each interpretation uniquely situated.
Recent scholarship, such as Homewood et al. (2021), reinforces that even identical data can evoke diverse interpretations due to individual sensory engagement. This insight underlines the limitation of digital systems that attempt to mimic human understanding solely through data analysis.
A central tension arises when contrasting these embodied approaches with recent developments in artificial intelligence, particularly with Large Language Models (LLMs) like ChatGPT. As Böhm et al. (2024) note, LLMs interpret semantic elements to generate contextually relevant responses. However, they inherently lack the bodily context that informs human sense-making. While LLMs may simulate a rudimentary Theory of Mind—as discussed by Leer et al. (2023)—they do so without the rich, embodied interactions that shape human emotional and cognitive experiences.
This divergence is further complicated by critiques from queer theory perspectives (Light et al., 2019), which argue that reducing human identity to mere metadata oversimplifies our lived realities. The model’s ability to 'remember' past interactions (OpenAI, 2024) offers an intriguing parallel to human memory, yet it falls short of replicating the physical and emotional context that underpins genuine human interactions.
From an Interaction Design standpoint, there is a compelling challenge: How can we design AI systems that recognize and integrate the embodied nuances of human experience? One potential direction is to refine the concept of Theory of Mind within AI, enabling systems to move beyond abstract metadata and engage with users as whole beings. This might involve developing “Embodied Artificial Intelligence” that interacts within physical spaces and leverages multimodal sensory inputs, rather than relying solely on textual data.
In summary, while current AI models demonstrate impressive capabilities in simulating cognitive processes, a critical tension remains between their abstract computational foundations and the deeply embodied nature of human experience. Narrowing this gap could not only enhance user interaction but also offer richer, more contextually aware AI systems that truly understand the nuances of being human.