Disentangle call-length labeling artifacts from age-related fraud differences

Determine whether the observed decline with apparent age in the proportion of calls labeled as outright fraud reflects a genuine difference in the fraud market or an artifact of shorter calls providing the automated holistic labeler with less sales-pitch context.

Background

The randomized identity experiment found that older-presenting identities received longer engagement from callers, while the proportion of calls labeled as outright fraud rather than legal lead-generation declined with apparent age. The paper reports this comparison as an exploratory, uncorrected result and identifies two possible interpretations: younger identities may genuinely attract more identity-harvesting fraud, or shorter calls to younger identities may prevent the automated labeler from observing enough sales-pitch context to classify them as spam rather than scam.

Resolving this issue would clarify whether the apparent age pattern in scam-versus-spam composition reflects caller behavior or a measurement artifact arising from the interaction between call duration and automated labeling.

References

The younger identities drew proportionally more student-loan, tax-relief and debt pitches---verticals the rubric codes as identity harvesting---while the older ones drew Medicare and insurance sales, which it codes as spam. We report this as a single uncorrected comparison among eleven, and note a competing explanation we cannot exclude: shorter calls give the labeler less sales-pitch context to see, so some of the difference may be an artifact of call length rather than a fact about the market.

— Anatomy of a Scam Call: What 10,000 real scam and spam calls reveal about how phone scammers operate  (2608.24127 - Traister et al., 25 Aug 2026) in Section 4.5, paragraph beginning “Two further patterns, flagged as exploratory”