Knowledge

Creativityis often viewed as one of the defining hallmarks of intelligence, yet the mechanisms underlying creative behavior remain poorly understood. This question has become increasingly important as artificial intelligence systems demonstrate remarkable abilities to generate novel text, images, music, and code, raising the possibility of more general forms of machine creativity and, ultimately, Artificial General Intelligence (AGI)—the capacity for flexible intelligence across diverse tasks and domains. Although creativity is frequently treated as a singular capability, evidence from cognitive neuroscience suggests that it instead emerges from interactions among multiple brain systems responsible for generative, evaluative, and exploratory processes regulated by coordination. In this paper, we propose the Generative–Evaluative–Exploratory–Coordination (GEEC) framework, a neuroscience-inspired functional framework that conceptualizes creativity as an emergent property of three primary creative processes—generation, evaluation, and exploration—whose interactions are dynamically regulated by a higher-order coordination process. Generative processes produce novel ideas through imagination and associative recombination; evaluative processes assess their utility and coherence; exploratory processes drive curiosity and learning under uncertainty; and coordination is conceptualized as a meta-process that regulates transitions among the other components. We review the neural mechanisms underlying these functions and examine how they have informed modern computational approaches, including generative models, reinforcement learning, meta-learning, neuro-symbolic systems, and cognitive architectures. Rather than viewing neuroscience as a roadmap to AGI, we argue that it provides a source of hypotheses about the computational principles that support creative behavior. The GEEC framework emphasizes functional correspondences rather than mechanistic equivalence and proposes that creativity is inherently multicomponent, emerging from dynamic interactions among generative, evaluative, and exploratory processes regulated through coordination. By synthesizing perspectives from neuroscience, cognitive science, and artificial intelligence, GEEC offers a unifying framework for understanding creativity across brains and machines.

Generative AI (genAI) is increasingly being integrated into critical appraisal and knowledge synthesis of research, with growing discourse on its value for qualitative research. Moreover, ChatGPT covers a major portion of the market share and is widely used by researchers and academics. However, a major gap remains in genAI’s ability to assess the reporting quality of qualitative research studies. This pilot study aimed to determine whether GPT-5 could reliably assess the reporting quality of health-related qualitative research studies. The Standards for Reporting Qualitative Research (SRQR) checklist was used as a framework for completing both human and AI chatbot assessments. The SRQR is a widely used validated checklist for enhancing the reporting of qualitative research. Overall, 68 peer-reviewed studies were selected for testing, comparing human SRQR assessments with GPT-5 assessments. Findings revealed a wide variability in agreement between human and GPT-5 assessments, with crude agreement ranging between 60 and 100% and interrater reliability ranging between 0 and 0.85 across checklist items. Simpler, objective checklist items such as study purpose were often correctly identified and had stronger agreement compared to nuanced items. The model struggled with the following SRQR items leading to an “Unclear” response: research design/paradigm (41%), researcher reflexivity (41%), data processing (27%), and trustworthiness (46%). These preliminary results suggest that GPT-5 may struggle with tasks requiring nuance and interpretation of qualitative research, which is rich in complex methods and differing worldviews. Currently, GPT-5 and other AI chatbots should be used with caution and alongside human oversight when assessing the reporting quality of qualitative research within public health and healthcare.

Census-style address records contain heterogeneous structures that must be decomposed into fine-grained fields before they can support linkage, geocoding, and administrative processing. This paper describes the design and implementation of a three-stage Planner–Manager–Worker pipeline using a locally deployed open-weight language model. The prototype includes address-specific extraction, same-model semantic consistency review, formatting, record-preserving chunking, and an exploratory bounded retry pathway. Although the prototype also implements workers for names, phone numbers, email addresses, Social Security numbers, and dates of birth, the quantitative evaluation is limited to address-component parsing. We conduct a controlled pilot using 700 author-generated synthetic records distributed across seven address categories. Each record was first created as a structured object and then rendered into text; the original structured object served as the evaluation reference. Across three repeated executions on the same fixed benchmark, the complete hierarchical configuration obtained a mean micro-averaged address-component exact-match score of 95.7%, compared with 80.9% for the evaluated monolithic LLM reference; the deterministic rule-based reference obtained 52.3%. The fixed prompts were not matched, and realized inference calls, dynamic token allocation, original batching, and reported-run segmentation could not be verified from retained artifacts. No component ablations were conducted. Consequently, the numerical difference cannot be attributed uniquely to planning, task coordination, specialization, semantic review, feedback, chunking, or multi-agent organization. The reference labels were not independently annotated, the benchmark does not estimate performance on naturally occurring records, and the retry pathway was invoked only three times. The contribution is therefore the specification and controlled pilot characterization of an on-premise address-parsing pipeline, rather than evidence of architectural causality, production readiness, broad PII-extraction accuracy, or real-record robustness.