Using Machine Learning to Make Assessment of Next Generation Science Standards–Based Three-Dimensional Science Scalable

• In Event: AERA Poster Session 8
  In Poster Session: Division H Section 3 Poster Session 2: Examining Assessment in Schools

Abstract

Assessing students on the NGSS will require the use of constructed response or composite items that are better able to elicit evidence of what students know and can do using science practices, cross cutting concepts in science, and disciplinary core ideas. However, accountability testing must be affordable or states will not implement it (Toch, 2006).

The NAP report on developing assessments for the Next Generation Science Standards (Pellegrino et al, 2014) stated that emerging technologies such as machine scoring would be necessary to meet the assessment needs of the ambitious NGSS. The purpose of this paper is to show that a machine-learning process can score science items holistically and that the process is scalable.

Authors

• Jay Thomas, ACT, Inc.

• Karen L. Draney, University of California - Berkeley

• Shruti Bathia, University of California - Berkeley