Reaching different conclusions from identical information because of how the options, risks or question were expressed. The framing is usually supplied by someone else — a customer, a policy document, a headline, or the person who wrote the tasking.
In intelligence analysis
- A tasking asked as "how serious is this threat?" and the same tasking asked as "is there any evidence of harm?" invite different answers from the same file.
- Gain and loss framings of a risk produce different recommendations: 90 per cent detection coverage and one in ten intrusions missed describe the same system.
- The customer's framing is often the strongest signal in the tasking, and it is the part least likely to be questioned — the analyst takes the question as given and answers it precisely, which is how a product can be accurate and still wrong.
- The same evidence, framed as a base-rate problem rather than a case problem, can reverse the judgement (see Base-Rate Neglect).
Countermeasure
- Rewrite the tasking in your own words before analysis begins, and state what the alternative framings would have produced.
- Express the same finding in both gain and loss terms, and check that the recommendation survives both.
Related
- Cognitive Bias — the taxonomy this bias sits within
- Base-Rate Neglect — the framing that most often reverses a judgement
- Bottom Line Up Front — the reader's framing instrument; what leads, frames
- Estimative Language — neutral phrasing reduces framing drift