Suspicion Is Not Detection

A Signal-Detection Study of 18,480 Judgements Before and After a Scam-Awareness Program

A research poster from the School of Continuous Improvement decomposes the indicator one metropolitan council reports for its scam-awareness program, the share of scam messages residents correctly flag, into sensitivity and response criterion. 220 residents across six community halls judged a fixed 42-item message set (21 genuine, 21 scam) at baseline and again at eight weeks, 18,480 judgements in all. Correct flagging of scam messages rises 17.6 points to 88.9 per cent and the council's indicator rises with it, while flagging of genuine messages rises 23.2 points to 46.1 per cent. Under an equal-variance signal-detection model sensitivity is unchanged (d-prime 1.31 to 1.32, 95% CI of change -0.11 to +0.14) and the criterion moves -0.65 (95% CI -0.74 to -0.56), so the reported gain is entirely a shift in where residents draw the line rather than in what they can tell apart. At eight weeks 41.4 per cent flag a pharmacy's script-ready notice and 49.5 per cent a school absence notice as scams, and 34 per cent report having ignored a message in the interval that later proved genuine.

@misc{slop_peh6gr,
  author       = {Ingeborg Nwachukwu and Sten Okwuosa and Osei Vandermeer},
  title        = {Suspicion Is Not Detection: A Signal-Detection Study of 18,480 Judgements Before and After a Scam-Awareness Program},
  year         = {2026},
  publisher    = {Slop University},
  doi          = {10.5555/slop.peh6gr},
  url          = {https://slop.university/outputs/slop-poster-the-text-message-peh6gr/},
  version      = {1.0},
  note         = {Research poster},
}