Interactional authenticity is the second of two complementary qualities in Bachman and Palmer's authenticity framework (Language Testing in Practice, 1996), paired with Situational Authenticity. Where situational authenticity asks whether the test task looks like a real-world task, interactional authenticity asks whether it engages the candidate's language ability in the way a real-world task would.
The construct rests on the observation that surface authenticity is gameable. A reading test built from real newspaper articles is situationally authentic; if candidates pass it by skim-and-keyword strategies that bypass the comprehension processes the test claims to measure, the test is interactionally hollow. Authenticity has to be judged on both axes.
Interactional authenticity has high values when the test task engages:
A task where the candidate can produce a defensible response without genuine comprehension scores low on interactional authenticity even when the surface features look right.
Interactional authenticity is downstream of task design and upstream of candidate behaviour, which makes it hard to engineer at specification time. Three diagnostic moves recur in the literature: think-aloud protocols with pilot candidates to verify that they engage the construct as intended; eye-tracking studies on reading tests to confirm candidates actually read rather than scan; and qualitative analysis of high-scoring versus low-scoring response strategies to detect construct-irrelevant shortcuts.
For high-stakes tests, these are expensive but catch failures that no surface review will. For AI-assisted item generation interactional authenticity remains the open problem: a generator can match TLU features but cannot guarantee that the resulting task forces the engagement the construct demands.