Academics from the University of Colorado and New Mexico State University have been testing a system they have developed called the Intelligent Essay Assessor (IEA).
The system, like other computerised essay-marking systems currently in development, uses a range of measures to analyse content and style, register concepts and spot plagiarism and mistakes.
The researchers tried out the system on thousands of essays from school to university level and found the difference between two human markers was usually greater than the difference between a marker and the computer.
They tested the computer on nearly 800 fictional stories written by middle school pupils.
The team concluded that such systems should now be used to give a second or third opinion on essays.
They said results suggested that "human graders are so unreliable that they can be matched by measuring almost anything that distinguishes good from bad essays".
Automatic Essay Assessment by Thomas K Landauer et al appears in Assessment in Education: principles, policy amp; practice Volume 10 Number 3