Choose your critic

My last post looked at the potential of a large language model such as ChatGPT to operate as a critic. Developing further on that theme, it occurred to me that such a review may not always be in step with the author’s target readership.

So, I presented ChatGPT with a 4,000 word book chapter titled “Urban actors” I am working on currently and asked for a critique on the basis of 5 different reviewer profiles: (1) an analytic philosopher, (2) a radical poststructuralist versed in the theories of Derrida and Foucault, (3) a neo-marxist reader, (4) someone versed in phenomenology, (5) a pragmatist more interested in practical applications than theory.

It’s not practical to include the chapter here, but this is a summary: In brief, the chapter addresses concepts of agency within urban environments and their relevance to AI large language models (LLMs) in urban contexts. Drawing on Actor-Network Theory (ANT), I proposed that agency extends beyond humans to include objects, technologies, and AI models, highlighting how LLMs act as agents by activating patterns of attention in text. I discussed multi-agent systems and their applications in urban planning, illustrating parallels between the computational agents in AI and elements in urban settings. I also looked at the concept of conversation-centric computing, reviewing the claims that AI agents might interact dynamically and solve complex problems collaboratively. Through practical demonstrations, I highlighted the potential and pitfalls of AI to simulate certain decision-making processes.

Here are the reviews generated by ChatGPT from the 5 different points of view.

(1) Analytic philosopher

This chapter presents a compelling exploration of agency, extending the notion beyond human actors to include objects, technologies, and artificial intelligence within urban contexts, drawing heavily on Bruno Latour’s actor-network theory. While the argument for considering non-human entities as agents is innovative, the chapter could benefit from a more rigorous examination of the philosophical implications of attributing agency to inanimate objects. The discussion on multi-agent systems and conversation-centric computing is insightful, yet it seems to skirt the deeper philosophical question of what constitutes agency and whether these AI systems genuinely possess it or merely simulate it. Moreover, while the analogy between urban planning methods and AI systems is intriguing, it would be strengthened by addressing potential ethical concerns regarding the deployment of such AI systems in urban environments, particularly in terms of autonomy and control. Overall, the chapter provokes thought on the evolving nature of agency in the digital age, but it would benefit from a more thorough philosophical inquiry into the essence and limits of agency.

(2) A radical poststructuralist versed in the theories of Derrida and Foucault

This chapter engages provocatively with the concept of agency by destabilizing the anthropocentric bias inherent in traditional understandings of action and influence. Drawing from Derrida’s deconstruction, the chapter effectively questions the logocentric privileging of human over non-human actors by including objects and AI as agents, thereby challenging the boundaries of agency and subjectivity. However, the text could further delve into the implications of this decentering for notions of power and control, as Foucault might suggest that the introduction of non-human agents reconfigures the power dynamics within urban spaces. The chapter flirts with the idea of AI as autonomous agents, yet it does not fully interrogate the discourse that constructs and legitimizes such technologies, nor does it critically assess the biopolitical ramifications of deploying AI in urban contexts. While the discussion on multi-agent systems and conversation-centric computing hints at a rhizomatic network of interactions reminiscent of Deleuze and Guattari, the analysis stops short of fully embracing the radical potential of these ideas to subvert hierarchical structures. Overall, the chapter initiates a valuable dialogue on the multiplicity of agency but could push further in its deconstructive and genealogical analysis to uncover the underlying assumptions and power relations at play.

(3) A neo-marxist reader

This chapter offers an intriguing exploration of agency that aligns with contemporary debates about the role of technology in capitalist society. The inclusion of non-human entities such as AI as agents reflects the increasing commodification and reification of technology, which Marxists might argue serves to obscure the social relations and labor processes underpinning their development and deployment. The chapter acknowledges the transformative potential of AI in urban contexts but does not sufficiently interrogate the capitalist imperatives driving its integration into everyday life. The author needs to analyze how these technologies reinforce existing power structures and exacerbate inequalities, potentially serving the interests of capital over those of the working class. Furthermore, the discussion on multi-agent systems and conversation-centric computing could benefit from a more critical examination of how these systems might be used to surveil and control populations, thus facilitating the reproduction of capitalist ideologies. While the chapter raises important questions about the nature of agency and interaction, it could more rigorously address the socio-economic conditions and class relations that shape and are shaped by technological advancements in urban environments.

(4) A phenomenologist who reads Husserl, Heidegger, Merleau-Ponty, etc

This chapter offers a compelling yet limited exploration of agency in urban environments, particularly when considering the relationship between technology and human experience. Phenomenology emphasizes the lived experience and the intentional structures of consciousness, focusing on how individuals perceive and engage with the world around them. While the chapter acknowledges the role of non-human entities as agents, it could benefit from a deeper examination of how these entities influence human perception and experience within urban spaces. The discussion on AI and multi-agent systems hints at the complex interactions between humans and technology, but it falls short of fully exploring the phenomenological implications of these interactions. For instance, how do these technological agents reshape our experience of space and time in the city? The chapter might also consider Heidegger’s notion of “Being-in-the-world,” which emphasizes the embeddedness of human existence within a broader context, including technology as a mode of revealing and concealing aspects of our being. By focusing more on the experiential and existential dimensions of technology in urban settings, the chapter could provide richer insights into how AI and agency intersect with the human lifeworld, potentially transforming our understanding of urban life and our place within it.

(5) A reader more interested in practical applications than theory

This chapter presents intriguing concepts about agency in urban environments and the potential role of AI in reshaping how we understand and interact with these spaces. However, there’s a lack of concrete examples and actionable insights into how these theories can be implemented in real-world urban planning and development. While the discussion on multi-agent systems and conversational AI suggests potential applications in urban management, traffic control, and collaborative planning, the chapter does not provide detailed case studies or evidence of successful implementations. The chapter could include specific examples of cities or projects where AI and agency theories have been effectively applied to solve urban challenges. Additionally, more emphasis could be placed on the technical aspects of integrating AI into urban systems, such as the infrastructure requirements, potential challenges, and measurable outcomes. By focusing more on these practical elements, the chapter could offer a more balanced perspective that bridges the gap between theoretical discourse and tangible urban improvements.

AI as helpful tutor

It looks as though my chapter is wanting from whatever framing you take, but then that’s part of the academic game. As any teacher or student knows, there’s always something to criticise. Everything can be made better. It’s often time constraints and deadlines that call a halt to the authoring and improving process.

From these 5 viewpoints it may seem as though you get it right if you choose the right critic. But that’s normal in academic production. The challenge of nominating a reviewer or examiner is to pick someone whose mindset corresponds in some way to that of the author, or someone who is at least sympathetic to your viewpoint. Another way of saying this: its about matching content to audience.

But I am framing this exercise in terms of helpful feedback and not assessment or examination. That also raises the prospect of using the AI conversationally rather than via a report. An effective critic hopes for a response rather than merely correction.

AI as critic warrants healthy skepticism. Knowing how an LLM operates and theories behind it may help. As any academic knows, you are entitled to critique your critic.

Note

Featured image is from ChatGPT: “Here is the image of a post-apocalyptic theatre stage to accompany your blog post. Feel free to use it as you see fit!”