The future of tests: Students can find a way to fool AI in exams

Artificial intelligence (AI) is no match for human ingenuity when put to the test, as has been made clear by the feat of high school students who have managed to best the algorithms used to conduct virtual learning assignments in the US.

Students who have been through the grind in public examinations in India — long before AI was even a glint in John McCarthy’s eye — know the importance of ‘key words’ to get marks.

It seems nothing has changed even with technology, as a Los Angeles schoolboy’s recent use of a “word salad”, appended in a random manner without even proper sentences, got him full marks on short questions in online assignments once he realised tests were being assessed by AI.


Many other students there then cited similar success with the same tactic. Indians cannot miss the uncanny similarity between AI’s predilection for awarding marks for just words instead of sentences and that of Indian examiners.

With the Covid-19 pandemic leading to an explosion in online learning, assignments and tests worldwide, the keyword trick of US students cannot have escaped the notice of Indians who have made a lucrative profession out of gaming examinations. The implication of Indian authorities succumbing to the temptation of roping in AI for this task in the future is obvious. Testing Indian resourcefulness is a tall order even for AI.