Email
Password
Remember meForgot password?
    Log in with Twitter

article imageArtificial intelligence used to mark exam essays

By Tim Sandle     Aug 2, 2017 in Technology
Artificial intelligence is increasingly being used to mark exam papers, taking the human marker out of the equation. This is not simple multiple choice scanning, but assessing essays too.
Artificial intelligence is being applied to new learning models to offer exciting possibilities for students. The technology also offers new ways to automate the teaching process and to reduce administration costs.
To boldly mark exams like no one has marked before
In the world of examination assessments, one of the first systems to mark examinations electronically was Assessment21. The artificial intelligence platform can mark exam papers electronically, for exams undertaken by students online. The software was developed at the University of Manchester with the aim of reducing the expenses associated with setting, administering and marking traditional paper exams.
An exam room
An exam room
Flickr user comedynose
The idea of a using a computer to scan multiple choice answers is not new and such software is commonplace. It is not dissimilar to those computers that drive regular surveys on Facebook and Twitter.
In education, computers have been programmed to scan the papers, recognize the possible right responses and compile the marks. What is different about ‘artificial intelligence’ for exam marking is that platforms can mark complex, open-ended questions designed to test students’ understanding. Intelligent software can learn to focus on key words in exam answers and to run these against a model answer. In addition, sentences can be scanned to check for plagiarism.
READ MORE: Increasing use of virtual reality in education
AI marking online
An alternative AI system is Gradescope, which was developed at UC Berkeley. The aim of this package was to grade papers to reduce the time required by teaching assistants to undertake this task.
The software evaluates online examinations. Since its inception the software has assessed 10 million answers to 100,000 questions across a wide-ranging college curricula. The artificial intelligence assesses three challenges: identifying question types, distinguishing between different written marks, and with recognizing handwriting (this latter part in in preparation for assessing future written exam answers). The designers, led by Sergey Karayev, state that exam marking time has fallen by 50 percent.
Students studying for an exam
Students studying for an exam
Albeiro Rodas
Marking down under
Artificial intelligence systems for exam marking have also been tested out in Australia, by the Australian Curriculum, Assessment and Reporting Authority. To prove effectiveness, the government body used supervised machine learning. For the trial, sample tests marked by humans were fed into an algorithm. The trial showed that the most effective algorithms can learn how to recognize quality responses by reverse-engineering scoring decisions. The aim is to begin the marking of several school based exam papers later in 2017 (students in years 3, 5, 7 and 9).
A spokesperson for the Australian examiners, Dr Rabinowitz said the artificial intelligence system would identify "sentence structure, vocabulary, grammar, right down to more mundane elements like spelling". In addition it can perform "as well as or even better than the teachers involved".
READ MORE: Digital technology can provide personalized learning for students
Teacher woes
According to IT News the use of algorithms for exam marking has met with some opposition from teachers. Teaching unions are worried that computer-based marking undermines the creative aspect of student performance and learning. It is uncertain whether the teacher opposition will slow-down the planned implementation. However this plays out, the benefits to schools and universities in reduced administration time are likely to propel this application of artificial technology forwards.
More about Education, Learning, Student, Artificial intelligence
 
Latest News
Top News