Delivered-To: bhaselton@mail-sttl.uswest.net Date: Fri, 23 Jun 2000 18:36:36 -0500 From: "Bennett Haselton" Sender: owner-peacefire-press@iain.com To: peacefire-press@iain.com Subject: Wired News reporter responds to plagiarism charges Reply-To: Bennett Haselton [sent to journalists on Peacefire's press contacts list] (this is unpleasant business and not exactly news, so I wouldn't blame you for skipping this message, but original post about the Wired article did get a lot of responses) The Wired News reporter, Declan McCullagh, who wrote the story about the BAIR filter based on our report at: http://www.wired.com/news/technology/0,1282,36923,00.html has responded to our plagiarism accusations in a message posted to a message board on WELL.com, copied below (the actual message boards are only available to WELL customers). I do believe, as Declan says, that he spent time verifying our results (that was why we offered him an exclusive, which I thought was only fair). However, his report at http://www.well.com/user/declan/bair/ did not uncover anything new that wasn't already covered in our report at http://www.peacefire.org/censorware/BAIR/ , which Declan read before he started his tests, and which included a section, "How you can duplicate these results in your own experiment". (In science, as in journalism, it's important for peers to verify your results, but the first discoverer is still supposed to get credit -- since it's a lot easier to verify someone else's discoveries, if you know exactly what to look for and what the results are going to be.) I think what Declan did was a disgrace; if I were an editor, I would consider firing him. Certainly we're not giving any more advance copies of our reports to Wired News -- which might be cheerful news for everyone else. (To respond to one of Declan's claims: we *did* test the AI image-blocking feature in the program, and it did block some images, just not very many (and none of the pornographic ones!). Declan said that when he tested BAIR, Exotrope told him to change an obscure setting that only high-level tech support knew about, which "toughened up" the AI and made it block a lot more images, but it was still random, blocking porn and non-porn with the same frequency. So we did test the AI, but only at the default level. What the Wired article did not say was that all of the images reported as "blocked" in the article, were only blocked after changing the secret program setting -- so this is not the configuration that BAIR customers would actually be using, which is how we tested the program.) I urge you to contact Declan at declan@well.com or 202 986 3455 to get both sides of the story if you want. I can also be reached at bennett@peacefire.org or 425 649 9024 if you feel you're getting contradictory information from both sides. (Unless you're like me, and you just want to move on.) Declan's posting: >>> >Briefly, Peacefire isn't telling you the full story. Bennett >Haselton tipped me off that BAIR had problems, but I did a weeklong >investigation on my own. I have all the relevant (and extensive) >correspondence with Exotrope, the Perl program I wrote, the >pornographic image list it used, the non-pornographic image list it >used, and the program's output at: > http://www.well.com/user/declan/bair/ > >In fact, Peacefire did *not* test the key feature: the program's >"AI" routines. The /1 setting they used with BAIR just uses the >blacklist. I used the /2 setting that turns on the image >recognition. Note their web page says images are unblocked if you >add ?foo to the URL, which wouldn't happen if it was doing image >recognition. >>> -Bennett bennett@peacefire.org http://www.peacefire.org (425) 649 9024