Machine-Scorable Complex Constructed-Response Quantitative Items: Agreement Between Expert System and Human Raters' Scores
- Bennett, Randy Elliot; Rock, Donald A.; Sebrechts, Marc M.
- Publication Year:
- Report Number:
- GREB-88-07aP (1991), RR-91-11
- ETS Research Report
- Document Type:
- Page Count:
- Subject/Key Words:
- Constructed Responses Expert Systems General Test (GRE) GIDE Graduate Record Examinations Board Quantitative Tests Scoring
This study evaluated agreement between expert system and human scores on 12 algebra word problems taken by GRE General Test examinees. Problems were drawn from three content classes (rate x time, work, and interest) and presented in four constructed-response formats (open-ended, goal specification, equation setup, and faulty solution). Agreement was evaluated for each item separately by comparing the system's scores to the mean scores taken across five content experts. Results showed the expert system to produce scores for all responses and to duplicate the judgments of raters with reasonable accuracy; the median of 12 correlations between the system and human scores was .88 and the largest average discrepancy was 1.2 on a 16-point scale. No obvious differences in scoring agreement between constructed-response formats or content classes emerged. Ideas are discussed for further research and development concerning the use of expert scoring systems in large-scale assessment programs and in interactive diagnostic assessment.