Can Mechanical Turk replace professional journalism?

A group of journalists and programmers have teamed up to look into whether the work of a professional journalist can be replicated by crowd sourcing through Amazon Mechanical Turk.

The experiment, hosted on the team’s, seeks to uncover how possible it is to provide accurate, informative and interesting journalism by dictating work based on an algorithm that is essentially the ‘robot boss’.

The experiment bears some resemblance to a recent New Scientist article which sought to produce an encyclopaedia entry about New York. 

What is interesting about this particular experiment is that it is trying to replicate the work of a professional, skilled job, through the use of many small instances of unskilled labour.

The idea is to produce a piece of journalism roughly 500 words in length, from an original, newly released scientific paper.

The task of churning out the 500 words will be split into a number of jobs.  Firstly the story can be worked out by the Turkers by asking a number on the site what are perceived to be the most interesting aspects of the story. 

This could then be combined to get an overall idea of the angle of the story – by getting Turkers to vote on which is the most interesting aspect.

Another task might be to use references from the paper to identify which researchers may be able to comment on the story.

An interface built by the team will receive this feedback and then proceed with other aspects of the process such as fact checking and editing.

The researchers say they will be surprised if everything goes to plan.

For example there are reports that Mechanical Turk has a rather bad track record recently, so the chances of getting even one legible response might be rather slim.

But the experiment does raise interesting questions around authorial intent.  How much would it matter if this process actually does work?  Would readers be happy to sift through something that was put together piecemeal from the work of an unknowing crowd of people? Or are they already thanks to the nature of sweatshop journalism?

According to a researcher there are already examples of similar processes in the industry: “Publishers have already automated some parts of the editorial process and out-sourced others. It’s probably only a matter of time before a publisher quietly conducts this experiment, if they haven’t already.

“We think it’s better to explore the pros and cons of this approach in public. We hope that chronicling our progress here will trigger debate about the future of journalism and of crowdsourcing.”