What I find frustrating about studies such as this one is that the methodology seems to indicate a near total lack of consideration for what language is, how it works, or how/why people use it. In other words, such studies rarely consider linguistics, rhetoric, (illocutionary) communication, intention, interpretation, and (perlocutionary) action. I just have a difficult time understanding how any statement can be isolated from its context without losing all but its base syntactical and semantic meanings (although linguists do this to help them study structure). How can one assess bias outside of the entire context of a message (i.e. rhetorical situation)?
“John Kerry is a no-good polecat,” said Senator Blowhard.
Is that a positive or negative reference to John Kerry? I contend that you cannot tell unless the statement is placed in its proper (social, political, economic, professional, historical, etc.) context. It appears to be negative. But proper textual analysis is necessary to make a final conclusion, and proper textual analysis includes consideration of the context of a message.
All that said: The type of analysis represented in this study has much currency in communications studies today–especially in academic journals about journalism and mass communication. I’m an odd duck–a rhetoric scholar from the English studies tradition. I’m not claiming that such analysis is worthless, only that I think it fails to account for the knowledge and scholarship established by complementary disciplines such as linguistics and rhetoric.
One negative I see coming from this study: Knee-jerk, partisan, anti-bias ranters may use this as proof of liberal bias–even in the absence of any human intention.
[A suggestion for the scholar: Run your test again in an off-election year and compare the results to the first study. Run at least two tests: one actor based like the first test and one issue based. The context problems will still exist. But you may discover some interesting things about Google's results in a non-election cycle. Same RQ and H1.]
One positive I see coming from this study: If the people at Google take it seriously (and they should), it might encourage them to refine their advanced news search so that users may choose what type(s) of media should be represented in search returns (in addition to names of specific news organizations currently allowed). That would be useful.