Think-aloud tests with information specialists of ai-systems Iris.ai and Yewno (Danish)

<p>This dataset contains the nine think-aloud tests conducted in April-June&nbsp;2021.&nbsp;</p> <p>Think-aloud tests were designed to test the&nbsp;extent an academic search could be conducted&nbsp;in the two AI-powered search systems, Iris.ai (https://Iris.ai/) and&nbsp;Yewno&nbsp;Discover&nbsp;(https://www.yewno.com/discover). Pilot tests and validation tests were undertaken with two independent reviewers in March 2021. The validity tests were based on the principles outlined in Kim 2009, testing the content, face and construct validity of the test instrument (appendix 1). Consequently, the grammar and consistency of the language were improved and questions were reframed before the final think aloud tests were conducted in April-June&nbsp;2021.</p> <p>&nbsp;</p> <p>Ten&nbsp;information specialists&nbsp;were invited to take part in the tests.&nbsp;One test person from the&nbsp;Yewno&nbsp;tests dropped out of the study, resulting in an overall drop-out rate of 10%.&nbsp;Accordingly, five think-aloud tests in Iris.ai and four in&nbsp;Yewno&nbsp;were held at&nbsp;the university libraries in Aarhus and Copenhagen.</p> <p>Two testers ran each test. One conducted the dialogue and guided the test person through the tasks set in the think-aloud test. The second, noted down the test persons behavior,&nbsp;humour&nbsp;and comments.&nbsp;All tests were recorded using Zoom, both audio, and screen were recorded as well as the test persons behaviour was observed. The recordings were saved to&nbsp;Edumedia&nbsp;for the duration of the project and destroyed thereafter.&nbsp;</p>