Introduction
Problem-solving is the cognitive process of transforming an initial state into an intended goal state using mental or symbolic operators. While Allen Newell and Herbert Simon's Information-Processing Paradigm historically conceptualized human cognition as analogous to computational programs, modern cognitive psychology highlights profound computational, structural, and epistemic divergences between biological minds and synthetic systems.
1. Processing Mechanisms: Heuristics vs. Algorithms
Human problem-solving operates under bounded rationality, constrained by limited working-memory capacity and computational resources. Consequently, human cognition relies extensively on Dual-Process Theory:
- System 1 (Heuristic Processing): Operates rapidly, automatically, and semi-unconsciously using shortcuts such as availability, representativeness, and affect heuristics. This allows humans to bypass cognitive overload in novel environments, albeit at the cost of cognitive biases.
- System 2 (Algorithmic Deliberation): Deploys slow, effortful, serial reasoning when heuristics encounter an impasse or when high precision is demanded.
- Computerised Processing: Computational architectures execute exhaustive deterministic algorithms, brute-force state-space traversals, or statistical machine learning models. Computers process multidimensional search spaces without experiencing fatigue or working-memory decay.
- Applied Contrast: In medical triage, a clinician relies on intuitive pattern matching and fast-and-frugal heuristics to quickly stabilize an emergency patient, whereas an expert diagnostic AI computes structured patient parameters against large datasets using Bayesian probabilistic weighting.
2. Embodiment, Epistemology, and the Affective Filter
Human cognition is inherently embodied and emotional, whereas computers manipulate formal symbolic tokens devoid of physical grounding.
- The Somatic Marker Hypothesis: Antonio Damasio demonstrated that human decision-making is fundamentally linked to affective bodily states. Patients with ventromedial prefrontal cortex (vmPFC) damage retain standard logical reasoning and normal IQ scores, yet fail catastrophically at everyday problem-solving because they lack visceral autonomic signals ("somatic markers") that unconsciously eliminate disadvantageous choices. Computers lack this affective guidance system entirely.
- Syntactic vs. Semantic Processing: John Searle's Chinese Room Argument establishes that computer systems operate purely at a syntactic level—manipulating symbols according to formal rules without semantic comprehension or intrinsic intentionality. Human thought, by contrast, possesses genuine semantic grounding.
- Embodied Tacit Knowledge: Hubert Dreyfus's phenomenological critique emphasized that human problem solvers utilize an unstructured, context-sensitive everyday background of coping skills. Computers struggle with this frame problem because formal programming requires implicit background contexts to be explicitly coded.
3. Metacognition, Insight, and Bridging Architectures
Differences also emerge in self-regulation and conceptual restructuring:
- Metacognitive Monitoring and Insight: Humans monitor their own cognitive operations (e.g., feeling-of-knowing, tip-of-the-tongue phenomena) and can restructure their mental representation of an obstacle to achieve sudden Gestalt insight, as historically demonstrated in Wolfgang Köhler's primate studies. In contrast, algorithmic systems adjust search trajectories strictly within predefined loss functions and parameter spaces.
- Unified Cognitive Architectures: Cognitive scientists have built computational models that attempt to replicate human psychological constraints rather than pure raw computation:
- ACT-R (Adaptive Control of Thought-Rational): Formulated by John Anderson, this framework integrates declarative memory with procedural production rules while incorporating human-like activation decays and latency limits.
- SOAR: Developed by John Laird, Allen Newell, and Paul Rosenbloom, SOAR implements the Problem Space Hypothesis, using sub-goaling and universal impasse resolution to simulate cognitive flexibility and human learning trajectories.
Conclusion
While computerised problem solving excels in computational throughput, precision, and high-dimensional optimization across well-defined problem spaces, human problem solving remains unmatched in handling unstructured ambiguity. Human problem solving leverages semantic intentionality, somatic emotional cues, and contextually grounded tacit knowledge, rendering human and machine problem-solving complementary paradigms rather than identical processes.