Identifying Promising Items: The Use of Crowdsourcing in the Development of Assessment Instruments

The psychometrically sound development of assessment instruments requires pilot testing of candidate items as a first step in gauging their quality, typically a time-consuming and costly effort. Crowdsourcing offers the opportunity for gathering data much more quickly and inexpensively than from most targeted populations. In a simulation of a pilot testing protocol, item parameters for 110 life science questions are estimated from 4,043 crowdsourced adult subjects and then compared with those from 20,937 middle school science students. In terms of item discrimination classification (high vs. low), classical test theory yields an acceptable level of agreement (C-statistic = 0.755); item response theory produces excellent results (C-statistic = 0.848). Item response theory also identifies potential anchor items without including any false positives (items with low discrimination in the targeted population). We conclude that the use of crowdsourcing subjects is a reasonable, efficient method for the identification of high-quality items for field testing and for the selection of anchor items to be used for test equating.