SATs are formal methods for countering cognitive bias and testing hypotheses — the modern complement to (and critique of) older methods like the Admiralty Rating system. Post-9/11 tradecraft stresses them; the CIA's Tradecraft Primer is the reference text.
Common SATs
- Key Assumptions Check — surface and question the assumptions underpinning a judgment.
- Analysis of Competing Hypotheses (ACH) — list competing hypotheses and test ALL evidence against EACH, rather than defending the best guess.
- Argument Maps / Mind Maps / Matrix — externalise reasoning structure.
- Chronologies / Timelines — order events to spot patterns.
- Starbursting, Cluster Brainstorming, Inconsistency Finder — generate and cross-check ideas.
Table of technique utility (from AIPIO 25, Kathy Pherson)
Different SATs help at different stages (identify data needed / analyse / publish visually). For example: Key Assumptions Check scores high on data identification and analysis, low on visual publication; Link Charts and Argument Maps score high across all three. Dedicated software tools now embed many of these SATs.
Why SATs but not replacement
- The Admiralty Rating system has no built-in mechanism for generating alternative hypotheses or highlighting bias — SATs fill this gap.
- Tradecraft valuation is as important as ever — an LLM can't reason from first principles; the analyst's application of SATs to both human sources and AI outputs is the modern skill.
- Blended model: use official grading as shorthand, but run assumption checks, red teaming and ACH on top to mitigate its limits.
Related
- Admiralty Rating — the older system SATs are designed to complement
- The Intelligence Cycle — SATs strengthen the analysis stage
- Estimative Language — articulate SAT conclusions with calibrated confidence
Open questions
- Which SATs best scale to cyber and OSINT data streams? (Noted as open in source.)