Agents in the Long Game of AI: Computational Cognitive Modeling for Trustworthy, Hybrid AI by Marjorie McShane, Sergei Nirenburg, and Jesse English.
Marjorie McShane is a cognitive scientist and computational linguist known for her work on cognitively inspired, trustworthy AI systems that can collaborate with humans in natural language. She is a professor in the Cognitive Science Department at Rensselaer Polytechnic Institute (RPI) and co-directs the Language-Endowed Intelligent Agents (LEIA) Lab.
Sergei holds a PhD in Linguistics from the Hebrew University of Jerusalem and an M.Sc. in Computational Linguistics from Kharkov State University. Jesse holds a PhD in computer science, with a focus on language understanding from Rensselaer Polytechnic Institute. He is a senior researcher in the LEIA Lab, leading the development of content-centric intelligent agent architecture.
Together they work in related areas: knowledge-rich natural language processing, cognitive architectures, and language-endowed intelligent agents,
AI has relied traditionally on machine learning. The position by the authors is that for over thirty years machine learning development, it is not an all-purpose solution to building human-like intelligent systems. One hope for overcoming this limitation is hybrid AI: that is, AI that combines ML with knowledge-based processing.
Agents are the key
The central message is for any reader who care about where AI is heading. The authors argue that if we want AI we can genuinely trust, we need systems that can understand, reason, and explain themselves—not just crunch data.
So, their core idea is to promote “hybrid AI” which is combining today’s powerful machine learning LLMs with explicit knowledge and reasoning systems. Instead of treating AI as a ‘black box,’ the authors suggest focusing on “agents” that have internal models to make decisions, and can tell you why they did what they did.
Another title, Trustworthy AI is book that I would recommend as a companion:
One of their most appealing themes is trust. The authors keep coming back to a central theme: Would you trust an AI doctor, tutor, or assistant if it couldn’t document its choices? They argue that truly trustworthy AI must be able to justify its actions in human terms, not just offer a confident?sounding answer. This makes the book feel less like a technical manual and more like a roadmap for more responsible, ethical AI in daily life.
That said, this book is not easy reading. Non-technical readers may find many sections highly complex. Someone without a background in AI might prefer to dip into selected chapters or just use the book as a way to understand the direction of “trustworthy AI” at a high level, rather than trying to ingest every technical detail.
In conclusion, this is a serious book that pushes beyond hype to ask what kind of AI society actually needs. For nontechnical readers who are patient and curious, this is a demanding but rewarding view into a future AI that actually might become more understandable, reliable, and collaborative.