🌐 Gemini — Conceptual Model Building for PUMA
Tool: Gemini (Google) Phase: Phase 1 — Research Step: 04 — Conceptual Synthesis
Prompt A — Thematic Cluster Map
Using the literature on AI agents for project management (2023–2026), synthesize the main research streams into a coherent thematic map. For each stream: (1) name and core question, (2) 3–5 representative papers, (3) dominant methodology, (4) key finding, (5) unresolved tensions with other streams. Show how PUMA bridges multiple streams simultaneously.
Prompt B — Theory Candidate Comparison
Identify 4–6 theoretical frameworks used in the literature for studying AI agents in project management tasks. For each: (a) name and authors, (b) dependent variable typically explained, (c) level of analysis (task/agent/system/organization), (d) central causal mechanism, (e) empirical predictions, (f) known limitations or critiques, (g) prior application to PM contexts and outcomes. Build a comparison matrix. Recommend the most appropriate framework for PUMA's DSR approach.
Prompt C — Cross-Paper Tension Analysis
Some papers in my PUMA corpus agree, others conflict on: (1) whether LLMs outperform classical ML for issue classification, (2) whether few-shot prompting beats fine-tuning for estimation tasks, (3) whether multi-agent architectures improve over single-agent baselines. For each tension: identify the methodological, dataset, or design reasons for disagreement. State which claims are strong enough to guide PUMA's experimental design and which require replication.
Prompt D — Research Roadmap Planning
I am moving from PUMA's literature phase to benchmark specification. Build a staged research plan using Keshav's Three-Pass Method as the reading backbone, covering: literature exploration → thematic clustering → deep reading → note synthesis → hypothesis generation → benchmark specification. For each stage: what to produce, best tool, artifact to save in Obsidian vault, and time estimate. Make the roadmap realistic for a one-semester TFG.
PUMA Relevance
The conceptual model generated here provides the visual/structural argument for PUMA Section 2. The theory comparison matrix justifies the DSR paradigm choice. The roadmap feeds directly into Sprint-02.