Recognizing Cognitive Patterns in AI Systems

Most conversations about AI focus on models. But the real innovation today is happening in how AI thinks, plans, acts, and improves — autonomously. This is where Agentic AI stands apart. Over the past year building agent systems, testing LangGraph, ReAct, ToT, Google A2A, MCP, and enterprise orchestration layers, one pattern has become clear: To build effective AI agents, you need more than prompts or tools — you need a cognitive operating system. Here is a simple, foundational framework called C-O-R-E-F, that captures how autonomous AI agents operate: C — Comprehend The agent understands the input, intent, and context. It reads prompts, data, documents, and knowledge bases to extract goals, constraints, and entities. O — Orchestrate It plans and reasons. The agent selects the best approach, breaks the goal into steps, and chooses the right strategy or chain-of-thought. R — Respond Execution happens. The agent calls tools, APIs, or systems, generates outputs, updates databases, schedules tasks, or creates content. E — Evaluate The agent checks its own work. It compares outputs, validates information, runs tests, or uses an LLM-as-a-judge to detect errors or inconsistencies. F — Fine-Tune The loop tightens. The agent refines its logic based on feedback or logs, learns from outcomes, and improves future performance. This cycle is not linear — it is iterative and continuous. Every advanced agent system eventually converges to this pattern, regardless of framework or model. If you're building agentic systems, start thinking in loops, feedback, and orchestration layers, not just responses. The future of AI belongs to those who design thinking systems, not just powerful models.

How were humans able to recognize that Newton's laws of motion govern both the flight of a bird and the motion of a pendulum? This ability to identify the same mathematical patterns across vastly different contexts lies at the heart of scientific discovery—whether studying the aerodynamics of bird wings or designing the blades of a wind turbine. Yet, AI systems often struggle to discern these deep structural similarities. 💡 The key may lie in mathematical isomorphisms—patterns that preserve their relationships regardless of context. For example, the same principles of fluid dynamics apply to blood flowing through arteries and air streaming over an airplane wing, or the motion of a molecule. This raises a fundamental question in artificial intelligence: how can we enable machines to understand the world through these invariant structures rather than surface features? 🚀 Our work introduces Graph-Aware Isomorphic Attention, improving how Transformers recognize patterns across domains. Drawing from category theory, models can learn unifying structural principles that describe phenomena as diverse as the hierarchical assembly of spider silk proteins and the compositional patterns in music. By making these deep similarities explicit, Isomorphic Attention enables AI to reason more like humans do—seeing past surface differences to grasp fundamental patterns that unite seemingly disparate fields. Through this lens, AI systems can learn and generalize, moving beyond superficial pattern matching to true structural understanding. The implications span from scientific discovery to engineering design, offering a new approach to artificial intelligence that mirrors how humans grasp the underlying unity of natural phenomena. Key insights include: ➡️ Graph Isomorphism Neural Networks (GINs): GIN-style aggregation ensures structurally distinct graphs map to distinct embeddings, improving generalization and avoiding relational pattern collapse. ➡️ Category Theory Perspective: Transformers as functors preserve structural relationships. Sparse-GIN refines attention into sparse adjacency matrices, unifying domain knowledge across tasks. ➡️ Information Bottleneck & Sparsification: Sparsity reduces overfitting by filtering irrelevant edges, aligning with natural systems. Sparse-GIN outperforms dense attention by focusing on crucial connections. ➡️ Hierarchical Representation Learning: GIN-Attention captures multiscale patterns, mirroring structures like spider silk. Nested GINs model local and global dependencies across fields. ➡️ Practical Impact: Sparse-GIN enables domain-specific fine-tuning atop pre-trained Transformer foundation models, reducing the need for full retraining. Paper: https://lnkd.in/e85wHyQY Code: https://lnkd.in/eQicTqHZ

'Large language models learn from the patterns in organizational communication and decision making. If certain groups have been described as less ready, less technical, or less aligned, LLMs can internalize that and repeat it in summaries, recommendations, or automated coaching. Resume screeners detect patterns in who was hired before. If an organization’s past hires reflect a narrow demographic, the system will assume that demographic signals “success.” Performance-scoring tools learn from old evaluations. If one group received harsher feedback or shorter reviews, the AI interprets that as a trend. Facial recognition systems misidentify darker-skinned individuals and women at significantly higher rates. The MIT Gender Shades study found error rates for darker-skinned women up to 34 percent compared to under 1 percent for lighter-skinned men. Predictive analytics tools learn from inconsistent or biased documentation. If one team over-documents one group and under-documents another, the algorithm will treat that imbalance as objective truth. None of these tools are neutral. They are mirrors. If the input is skewed, the output is too. According to Harvard Business Review, AI systems “tend to calcify inequity” when they learn from historical data without oversight. Microsoft’s Responsible AI team also warns that LLMs reproduce patterns of gender, racial, and cultural bias embedded in their training sets. And NIST’s AI Risk Management Framework states plainly that organizations must first understand their own biases before evaluating the fairness of their AI tools. The message is consistent across institutions. AI amplifies the culture it learns from. Bias-driven AI rarely appears as a dramatic failure. It shows up in subtle ways. An employee is repeatedly passed over for advancement even though their performance is strong. Another receives more automated corrections or warnings than peers with similar work patterns. Hiring pipelines become less diverse. A feedback model downplays certain communication styles while praising others. Talent feels invisible even when the system claims to be objective. Leaders assume the technology is fair because it is technical. But the system is only reflecting what it learned from the humans who built it and the patterns it was trained on. AI does not invent inequality. It repeats it at scale. And scale makes bias harder to see and even harder to unwind.' Cass Cooper, MHR CRN https://lnkd.in/e_CXSdRE

Agentic AI Design Patterns are emerging as the backbone of real-world, production-grade AI systems, and this is gold from Andrew Ng Most current LLM applications are linear: prompt → output. But real-world autonomy demands more. It requires agents that can reflect, adapt, plan, and collaborate, over extended tasks and in dynamic environments. That’s where the RTPM framework comes in. It's a design blueprint for building scalable agentic systems: ➡️ Reflection ➡️ Tool-Use ➡️ Planning ➡️ Multi-Agent Collaboration Let’s unpack each one from a systems engineering perspective: 🔁 1. Reflection This is the agent’s ability to perform self-evaluation after each action. It's not just post-hoc logging—it's part of the control loop. Agents ask: → Was the subtask successful? → Did the tool/API return the expected structure or value? → Is the plan still valid given current memory state? Techniques include: → Internal scoring functions → Critic models trained on trajectory outcomes → Reasoning chains that validate step outputs Without reflection, agents remain brittle, but with it, they become self-correcting systems. 🛠 2. Tool-Use LLMs alone can’t interface with the world. Tool-use enables agents to execute code, perform retrieval, query databases, call APIs, and trigger external workflows. Tool-use design involves: → Function calling or JSON schema execution (OpenAI, Fireworks AI, LangChain, etc.) → Grounding outputs into structured results (e.g., SQL, Python, REST) → Chaining results into subsequent reasoning steps This is how you move from "text generators" to capability-driven agents. 📊 3. Planning Planning is the core of long-horizon task execution. Agents must: → Decompose high-level goals into atomic steps → Sequence tasks based on constraints and dependencies → Update plans reactively when intermediate states deviate Design patterns here include: → Chain-of-thought with memory rehydration → Execution DAGs or LangGraph flows → Priority queues and re-entrant agents Planning separates short-term LLM chains from persistent agentic workflows. 🤖 4. Multi-Agent Collaboration As task complexity grows, specialization becomes essential. Multi-agent systems allow modularity, separation of concerns, and distributed execution. This involves: → Specialized agents: planner, retriever, executor, validator → Communication protocols: Model Context Protocol (MCP), A2A messaging → Shared context: via centralized memory, vector DBs, or message buses This mirrors multi-threaded systems in software—except now the "threads" are intelligent and autonomous. Agentic Design ≠ monolithic LLM chains. It’s about constructing layered systems with runtime feedback, external execution, memory-aware planning, and collaborative autonomy. Here is a deep-dive blog is you would like to learn more: https://lnkd.in/dKhi_n7M

We have foundation models for language, images, and code. But what about the actual knowledge itself - the interconnected facts about the world that power reasoning? This is the goal of Knowledge Graph Foundation Models (KGFMs). Think of them as AI cartographers. Their job isn't to generate text or pictures, but to learn the invisible map of relationships between things - like how Intel, CPU, and supply chain connect. The real test is generalization: can a model trained on a graph of finance terms correctly navigate a new, unseen graph of tech manufacturing, just by recognizing that provide in the first graph and supply in the second play the same structural role? A groundbreaking new paper reveals a crucial bottleneck in how these models learn. It turns out that today's leading KGFMs, like ULTRA, learn by analyzing only the simplest possible patterns - specifically, how pairs of relations interact. This is akin to trying to understand a complex novel by only looking at two-word phrases. You get connections, but you miss the plot, the subplots, and the deeper narrative structure. The researchers introduce a powerful new framework, aptly named MOTIF, that breaks this limitation. MOTIF allows models to learn from richer, higher-order patterns - like how triples or even larger groups of relations interact. This is the leap from analyzing word pairs to understanding full sentences and paragraphs. Theoretically, they prove this isn't just a tweak; using these richer patterns gives the model strictly more reasoning power, allowing it to distinguish between complex relational scenarios that were previously indistinguishable. The results speak for themselves. Across a massive suite of 54 real-world knowledge graphs - from biology to social networks - models equipped with MOTIF's richer motifs consistently outperform the previous state-of-the-art. On particularly tricky datasets with conflicting patterns, the improvement can be dramatic, like a 45% boost in accuracy. This work is a paradigm shift. It suggests that the next leap in AI's ability to reason over knowledge might not come from simply scaling up data, but from fundamentally enriching the "vocabulary of patterns" the model has access to from the start. We are moving from models that can talk about knowledge to architectures designed to truly understand its structure. Link to full-length paper: https://lnkd.in/gPPD8f2W #AI #MachineLearning #KnowledgeGraph #FoundationModels #ArtificialIntelligence #Research #GraphNeuralNetworks #KnowledgeGraphFoundationModels

The CHI Tools for Thought Workshop brought together the world's top researchers on computer-human interaction. These are some of their extremely useful findings on the perils and potential of GenAI. 🧠 GenAI reshapes critical thinking. People often shift from active seeking to passive consumption of AI outputs, especially when trust in AI is high or domain confidence is low. This can lead to reduced reflection, overreliance, and homogenized thinking. 📚 Novices benefit least—and may be harmed. Underprepared or underconfident students often misuse GenAI, asking vague questions and following poor suggestions. These users show less critical thinking and get worse results than peers with more knowledge. 🎨 Creative workflows risk fixation. GenAI can accelerate design work but also encourages "tweaking" over exploration. Its high-fidelity outputs may fixate user thinking and reduce originality unless consciously countered. 💼 Experts want support, not substitution. Professionals embrace GenAI for routine tasks but avoid it for nuanced decisions. They value systems that augment rather than override their workflows, preserving agency and deep work. 🌱 Motivation and identity are at stake. GenAI may undercut intrinsic motivation by replacing meaningful mental effort. In creative fields, people resist AI replacing core contributions that define their professional identity. 🔧 Scaffolding beats full automation. Process-oriented AI—supporting steps like planning or schema formation—helps users better than fully automated systems. It’s most effective for complex tasks and learning goals. 💡 Cognitive friction can be a feature. AI systems that challenge users—by prompting reflection or surfacing ambiguity—can enhance thinking. But in productivity contexts, their value must be clearly evident to gain adoption. 🌀 Representation shapes understanding. Translating information across modalities or levels of abstraction can aid cognition. Examples include turning text into visuals or informal ideas into formal code. 🎭 Emotions and intuition can be augmented too. GenAI can boost ‘System 1’ processes like emotion and intuition to support cognitive outcomes. Examples include surreal stimuli to spark creativity, or personalization to increase motivation and reduce anxiety. 🛠️ Interfaces direct thought. Moving beyond text prompts, designs like direct manipulation or AI output previews can clarify user intent and reduce effort. But they might also reduce opportunities for deep reflection. 🔗 Workflow integration is key. GenAI’s real power comes when it supports entire workflows—not just tasks—especially in collaborative settings. Systems must adapt to roles, expertise, and context to augment rather than disrupt cognition. 📏 Better theories and measures are needed. Current frameworks help, but new constructs are needed to study how GenAI affects thinking. Reliable metrics will be crucial for assessing long-term cognitive impacts.

Most AI systems today rely on a single cognitive mechanism: procedural memory. That’s the kind of memory involved in learning repeatable patterns — how to ride a bike, follow a recipe, or autocomplete a sentence. It’s also the dominant architecture behind LLMs: self-attention over statistical embeddings. That explains a lot about LLM strengths as well as their failures. LLMs do well in what psychologist Robin Hogarth called “kind” environments — stable, predictable domains where the same actions reliably lead to the same outcomes. But they tend to fail in “wicked” environments—settings where the rules shift, feedback is delayed, and the right answer depends on context that isn’t explicitly stated. In those environments, procedural strategies break down. Humans rely on other mechanisms instead: semantic memory for organizing abstract knowledge, associative learning for recognizing useful patterns, episodic memory for recalling prior experiences. LLMs don’t have those. So they: ➡️ miss abstract relationships between ideas ➡️ fail to generalize across context ➡️ lose track of evolving goals ➡️ don’t build up any durable sense of what works and what doesn’t This isn’t a matter of more data or better training. It’s an architectural limitation. At Aampe, we’ve had to grapple with these gaps directly — because customer engagement is a wicked learning environment. That pushed us to move beyond purely procedural systems and build machinery that can form and adapt conceptual associations over time. Working on these problems has made me uneasy about how singular LLM cognition really is. If one mechanism were enough, evolution wouldn't have given us several.

If you have been reading my recent posts, you might have noticed a pattern. The Apple 'Illusion of Thinking' paper. Knuth's combinatorics test. The 'LLMs Can't Jump' abductive reasoning study. The SWE-CI maintenance benchmark. And now a Caltech-Stanford survey on reasoning failures. I keep writing about how these systems break. This is not an accident. Understanding the fail state is, for me, the single most important requirement in AI right now. Personally, because I find that failures reveal more about a system's nature than its successes ever do. Professionally, because when you build AI that makes important decisions for large institutions, knowing exactly where and how reasoning collapses is not optional. This piece is about a finding that stopped me cold: LLMs fail in ways that are nearly indistinguishable from human cognitive biases. Confirmation bias. Anchoring. Inhibitory control failures. Same behavioral signatures. Completely different causes. And that distinction changes everything about how you build, evaluate, and trust these systems. https://lnkd.in/gvXEatSt