The Sunday Times, 2nd November 2008, Front Page Headline:
Republicans try to use Oxford don to smear Barack Obama
Sarah Baxter, Washington
The Republicans have made a last-minute attempt to prevent Barack Obama's ascent to the White House by trying to recruit an Oxford academic to "prove" that his autobiography was ghostwritten by a former terrorist.
With two days before the election, Obama is poised to become America's first black president, according to polls showing he has an average six-point lead over John McCain, his Republican opponent.
Dr Peter Millican, a philosophy don at Hertford College, Oxford, has devised a computer software program that can detect when works are by the same author by comparing favourite words and phrases.
He was contacted last weekend and offered $10,000 (£6,200) to assess alleged similarities between Obama's bestseller, Dreams from My Father, and Fugitive Days, a memoir by William Ayers.
Ayers, now a university professor in Chicago, co-founded the Weathermen, a radical 1960s underground group that bombed government buildings in Washington and New York. The Republicans accuse Obama of "palling around" with him.
The offer to Millican to prove that Ayers wrote Obama's book was made by Robert Fox, a California businessman and brother-in-law of Chris Cannon, a Republican congressman from Utah. He hoped to corroborate a theory advanced by Jack Cashill, an American writer.
Fox and Cannon each suggested to The Sunday Times that the other had taken the initiative.
Cannon said that he merely recommended computer testing of the books. He doubted whether Obama wrote his autobiography, adding: "If Ayers was the author, that would be interesting."
Fox said he had hoped that Cannon would raise the $10,000 to run a computer test. "It was Congressman Cannon who initially pointed me in that direction and, from our conversation, I thought he might be able to find someone [to raise the $10,000]."
He believed that if "proof" of Ayers's involvement was provided by an Oxford academic it would be political dynamite.
Fox contacted Millican, who said: "He was entirely upfront about this. He offered me $10,000 and sent me electronic versions of the text from both books."
Millican took a preliminary look and found the charges "very implausible". A deal was agreed for more detailed research but when Millican said the results had to be made public, even if no link to Ayers was proved, interest waned.
Millican said: "I thought it was extremely unlikely that we would get a positive result. It is the sort of thing where people make claims after seeing a few crude similarities and go overboard on them." He said Fox gave him the impression that Cannon had got "cold feet about it being seen to be funded by the Republicans".
Cannon insisted, however, that he was not interested in making an issue of Obama's memoir "even if it were scientifically proven" to be someone else's work.
Obama said this weekend that the campaign would "get nasty" in its closing days. Last night he was forced to deny that he knew a Kenyan aunt was living illegally in the US. Zeituni Onyango, half sister of his late father, lost a bid for asylum in 2004. Obama said he had no knowledge of her status, but that the law should be obeyed.
The Sunday Times, 2nd November 2008, Feature, pages 16-17:
How they tried to tarnish Barack Obama
Peter Millican reveals how he was drawn into a plot to link the Democrat to a former radical
Last Sunday I received an urgent call from Bob, a man close to a Republican congressman in the American west. He wanted to enlist my services to prove a scandalous allegation against Barack Obama, which would surely affect his prospects in the forthcoming election. Namely, that his famous 1995 memoir, Dreams from My Father, on which so much of his reputation was built, was in fact written largely by Bill Ayers, a Vietnam-era domestic terrorist.
Many Republican attacks, in the mainstream media, websites and blogs, have focused on this connection, described by Sarah Palin as Obama's "palling around with terrorists". The new scandal held the promise not only of proving that Obama was a pal of Ayers, but even that he was, in a sense, Ayers's creation, groomed for high office by his mentor's ghostwriting hand.
Viewed from Britain all this seems like fantasy, but in some US circles it is taken very seriously indeed. If the charge against Obama could be made to stick, with objective evidence of Ayers's involvement in his famous book, then perhaps it could sway significant numbers of voters back to the Republican side.
So where do I come in, an Oxford philosopher and 18th-century scholar? Well, until three years ago I taught at Leeds University in a job that combined philosophy with computing. While there I wrote several software systems (available online) designed to make the study of computing more attractive and relevant for students in the humanities.
One of these, called Signature, performs "stylometric" analysis for literary detection. The idea is that by analysing different texts one can try to identify the distinctive authorial fingerprint or stylistic "signature" of each writer and thus decide questions of disputed authorship. Did St Paul write the Epistle to the Hebrews? Did Shakespeare write Edward III? Who wrote various anonymous political works, such as the Federalist Papers of 1787 or the 1996 novel Primary Colors?
Investigators have hoped to answer all of these questions – and many others – by such means and have sometimes been very successful (although more often wildly over-optimistic and naively uncritical).
My Signature system acquired some publicity this year through its involvement in a heated debate about Coleridge's alleged authorship of a translation of Goethe's Faust. So some Republicans were keen to make use of my expertise to help them in their quest to unmask Ayers as the hidden puppet master behind the Obama of 1995.
The person who came up with this strange theory is Jack Cashill, an American author who claimed to find striking similarities between Dreams from My Father and Ayers's 2001 memoir Fugitive Days.
The trouble with these sorts of claims is that they are far too easy to make: take any two substantial memoirs from the same era and you are likely to be able to pick out a fair number of passages that have some similarities. Unless the similarities are really close (and they weren't), just listing them makes no case at all, even if it might be enough to persuade some readers.
Cashill and friends – who were convinced but aware that more evidence would be needed to convince others – enlisted teams of analysts to try to give the theory a solid statistical basis. All of these analyses supposedly delivered positive results, but they seem badly flawed.
One pair of university engineers tested frequencies of words in the two books against a "random control"; but the book chosen for this role was a 1919 road novel which just happened to be easily available on the web. Hardly surprising, then, that the two contemporary memoirs came out relatively similar: they were always likely to be closer to each other than to a novel written 80 years before.
Another "team" – consisting of a private engineering consultant – used my own Signature software but didn't get beyond the first primitive test for word length frequencies (proportions of one-letter, two-letter, three-letter words, etc).
Although he did rigorous statistics with more appropriate "controls", his results were rather weak. But even if they had been stronger, using such limited data is like trying to argue that two houses were constructed by the same builder purely on the basis that they contain roughly similar proportions of different bricks.
Finally, the third analysis that I have seen used far more sophisticated measures, built into a proprietary software system for helping aspiring writers to develop their style. But it records no "control" measurements at all, so the results produced are impossible to assess.
The author of the analysis describes some of his results as "striking", but this looks very exaggerated to me. Without any comparable statistics involving other texts, we have no way of assessing their true significance. So his claim to have made "a strong case for the likelihood that the author of Fugitive Days ghostwrote Dreams from My Father" is completely unsubstantiated.
Bob – the man who brought me into all this – seemed sincerely interested in getting to the truth about Cashill's dramatic allegation. He supplied me with the relevant texts and a number of appropriate "controls".
Some preliminary tests, using various data measures and a range of powerful statistical facilities that were recently added to Signature, indicated nothing that would give Obama any cause for concern. So I felt that any analysis I did would be far more likely to put an end to the story than to substantiate it, by providing objective data against what looked like partisan allegations.
The Republicans were apparently keen to press for a full-scale investigation, which would take a good deal of my time but for which they were prepared to pay through Oxford University Consulting's personal consultancy arrangements.
Oxford University Consulting, on my behalf, insisted quite properly that any such arrangement would have to be agreed before the results were known: there could be no question of carrying out an analysis that would be paid for only if the results came out in their favour. And I insisted that the analysis, once produced, would have to be in the public domain and thus made available to the Democrats also.
Having got to this stage, with texts and controls carefully prepared and special facilities added to Signature for the purpose, my little adventure into US politics ended. I was left with the impression that payment for propaganda was fine; but payment for objective research was quite a different matter.
Maybe one day I'll go back and do the analysis in detail, but I doubt it. I would rather spend my time on serious research questions than on improbable theories proposed with negligible support.
For more technical detail follow links on Dreams from My Father at the author's website, www.philocomp.net.