Active reading turns passive information into actionable knowledge through deliberate questioning
Reading a paper without a framework produces passive familiarity — the feeling of having understood something — without the ability to use, critique, or extend it. Active reading converts reading into thinking.
The Two Modes of Research Reading
Passive reading: following the text, absorbing facts, noting interesting claims. Produces recognition (“I’ve seen this before”) but not retrieval under pressure (“I need this idea now”).
Active reading: maintaining questions while reading that force the reader to connect, evaluate, and generate. Produces flexible, usable knowledge.
The MIT AI Lab WP 316 three questions (Q1/Q2/Q3) operationalise active reading. Keshav’s Three-Pass Method provides the structural scaffolding within which they operate.
The Cognitive Offloading Risk
Risk of Delegating to AI
Modern AI tools (Claude, Perplexity, NotebookLM) make it easy to ask a paper’s questions to an AI rather than engaging with the text directly. This is efficient but dangerous: the AI produces a summary; the researcher never has to struggle with the text; the knowledge does not transfer.
Cognitive offloading (delegating thinking to AI) degrades the researcher’s own synthesis capability over time. The Q1/Q2/Q3 framework must be applied by the researcher, not delegated. AI can verify or challenge your Q1/Q2/Q3 answers — but the answers must come first from you.
PUMA Application
In PUMA’s SLR (≥40 papers, OE1), the reading protocol is:
- AI tools (Perplexity, Consensus, Elicit) for Pass 1 discovery — finding relevant papers
- Manual reading for Pass 2 comprehension — grasping the argument
- Q1/Q2/Q3 for Pass 3 integration — connecting to PUMA’s own research
This division preserves the researcher’s cognitive engagement where it matters (Passes 2–3) while using AI efficiency where it does not compromise understanding (Pass 1 discovery).
References
- MIT AI Lab. (1988). WP 316. LN-MITAILab-WP316-HowToDoResearch
- Keshav, S. (2007). Keshav-Reading-Log