A reader at The Robert Gordon University in Scotland tried to access my site from campus and received this message.
Now, presuming this is not a mistake, I am very proud of this achievement. It is my first distinction in what I hope will be a long and lasting line of them.
But then it’s probably a mistake. That’s statistics for you. Doing as much harm as good.
RGU apparently uses the firm Bloxx to filter webpages, and it’s probably their “Tru-View technology” (well, they have to call it something) which locked me out. I’m hedging because I don’t know for sure.
It could be that somebody at the university flagged me as dangerously violent, or that yet another statistical algorithm has gone off the rails. I’m hoping for the former, because that would make me a kind of celebrity, but I’m guessing it’s the latter, because I know how dicey classification algorithms are.
In this case, one of four things can happen: (1) the software can correctly let a benign site be displayed to the sensitive RGU readers; (2) it can properly block sites university administrators deem unacceptable; (3) it can improperly block harmless sites like mine; and (4) it can accidentally let a horrible site (say, the New York Time) slip by.
Generally, modern enlightened peoples being what they are, the costs of (4) are thought to be much higher than the costs of (3). Think of the line of “shocked” and “offended” students out the president’s door if the Scottish equivalent of (for example) HotAir were let in. But if somebody can’t see a site? Well, out of sight, out of mind.
Thus, the knobs in these algorithms are often tuned to the Better-Safe-Than-Sorry setting. Plus, add to that the impossibility—as in impossibility—of designing any statistical algorithm that can classify perfectly, and lots of mistakes will happen.
I wrote Bloxx and asked them to remove me from the stern eye of their filter. I’m sure they will if this was merely a bad algorithm. Of course, if somebody reported my site as violent, then that’s something very different. Stay tuned for updates.
Well on your title page you have men in Gangster outfits (liberal software does not distinguish cowardly mobsters from Police Detectives or ruggedly individual Private Eyes, all masculine archetypes being equally bad…), a skull, and the ultimate horror a lit cigarette being smoked, will no one think of the children… Face it bub you’ve brought this on yourself, I’d check the surrounding airspace for attack drones if I were you…
Mike,
Oh, I forgot the deep, pathological horror Brit puritans have of smoking!
…had to be the “Death Panels” group that got snagged in the look-up table.
Luke,
Death panels are a fine thing. As long as I’m in charge of them.
It’s the hat, Briggs, the hat. Makes you look “connected”.
Briggs,
I’m guessing that RGU is actually turning the knobs, not Bloxx. RGU has it in for you. Likely some term or terms (or phrases, etc) used are on the RGU banned list — you know universities, open minds, and such.
Can I ask if you intend to go for the full set of “Illegal, pornography-Adult, violence, hacking, racist”?
Also, would “pornography-Adult” be a subset of “pornography” from which one could infer that other types of this material would be permissible?
I spent a number of years writing software for the purpose of “internet policy management” at Nova Chemicals. My system did not block sites. All it did was send the “potential abuser” an email list of URLs they had visited that “seemed inappropriate” This gave them a chance to reform or to complain if they felt their activity was legitimate. A second incident within a month or so would result in another warning cc’d to human resources.
The final “algorithm” was very simple and resulted in almost no false positives. False negatives were also rare since over time a real porn surfer would always do something to get a letter.
Robert Graham’s paper, “A plan for Spam” got me interested in the Bayesian approach to quantification of the probability that “abuse” was occurring. This continues to be a fascinating study that led to Jayne’s book and your website.
Jeez, I got a PhD from RGU
I feel embarrassed all round.
Probably to to with the gun and gay topics
Why is this university limiting access at all?
The NHS computer systems do this periodically, it then goes away after a few hours. It’s a glitch, nothing more.
Although I am surprised any Uni has monitoring on the web. The ones I have worked at were a free for all.
Almost all UK universities have some kind of web nanny system to try to prevent “undesirable” material from being accessed via their servers. It was brought in over the past ten years or so as the capacity to render images and access streamed video of undesirable content in real time became commonplace on pcs and universities got frightened of the legal and publicity consequences of letting students see that sort of thing.
I am told that most of the content filtered is pornographic or politically sensitive (racist sites, etc) but instances like this do suggest that the net is cast quite widely and one would hope that the blocking of this blog is simply a mistake and not due to some political agenda.
P(B) errors?