Robots have historically been accused of taking jobs away from humans, so a software bot that hands out tasks for people to carry out may make a refreshing change.
Suggestbot, developed by Dan Cosley at Cornell University in Ithaca, New York, and colleagues, could help online communities such as Wikipedia and Slashdot distribute editing tasks. Such organisations rely on members to add and edit content but, as work piles up, it can be hard even for dedicated users to pick out appropriate tasks.
Suggestbot links tasks with people鈥檚 interests. It can comb through thousands of Wikipedia articles with a 鈥渘eeds work鈥� tag and compare them to the list of previously edited articles on a user鈥檚 profile, looking for similar articles. To test if this could increase productivity, Cosley studied the work of 91 Wikipedia editors, who had collectively requested 3094 tasks. He tested three versions of the Suggestbot algorithm 鈥� one compared titles of 鈥渘eedy鈥� articles with those in the editor鈥檚 profile, the second paired people with tasks popular with editors with a similar history, and the third looked for links between needy articles and those in an editor鈥檚 profile.
Advertisement
Each of the algorithms was equally likely to be used to generate a task, as was a random task generator. Two weeks after the tasks were handed out, just 1 per cent of the randomly chosen articles had been edited, compared to 3.4 to 4.3 per cent of the articles selected by the intelligent task recommenders. Cosley believes combining the three strategies could increase this further.
