Automating Performance-Based Tests

Going by the complexity of the issues that need to be addressed in designing performance-based tests, it is clear that automating the procedure is no easy task. The sets of tasks that comprise a performance-based test have to be chosen carefully in order to tackle the design issues mentioned. Moreover, automating the procedure imposes another stringent requirement for the design of the test. In this section, we summarize what we need to keep in mind while designing an automated performance-based test.

We have seen that in order to automate a performance-based test, we need to identify a set of tasks that all lead to the solution of a fairly complex problem. For the testing software to be able to determine whether a student has completed any particular task, the end of the task should be accompanied by a definite change in the system. The testing software can track this change in the system, to determine whether the student has completed the task. Indeed, a similar condition applies to every aspect of the problem-solving activity that we wish to test. In this case, a set of changes in the system can indicate that the student has the desired competency.

Such tracking is used widely by computer game manufacturers, where the evidence of a game player’s competency is tracked by the system, and the game player is taken to the next ‘level’ of the game.

In summary, the following should be kept in mind as we design a performance-based test.

  • Each performance task/problem that is used in the test should be clearly defined in terms of performance standards not only for the end result but also for the strategies used in various stages of process.
  • A user need not always end up accomplishing the task; hence it is important to identify important milestones that the test taker reaches while solving the problem.
  • Having defined the possible strategies, the process and milestones, the selection of tasks that comprise a test should allow the design of good rubrics for scoring.
  • Every aspect of the problem-solving activity that we wish to test has to lead to a set of changes in the system, so that the testing software can collect evidence of the student’s competency.