This is mainly a question of ethics.
Dr. Plait, how would you handle the following hypothetical scenario? You have a press conference, at which you debunk some sort of Bad Astronomy. Hours later, you discover that the data that you presented at the press conference were bogus. Someone on your staff transcribed them incorrectly. The true data, though they don't exactly support the BA that you were trying to debunk, are much less harmful to it than the bogus data that you presented. What will you do? Will you call another press conference, and apologize for the mistake? If you do this, your retraction may bolster the same BA that you were trying to debunk. Will you choose to skip the apology, and just keep quiet, and hope that no one notices that your data were bogus? This would surely be a violation of scientific ethics, and in the long run this would probably do more harm than good. It's a painful decision to make, but I think we all can agree which choice would be the right choice.
While browsing recently, I came across a number of angry web pages which argued that a certain image of the "face" on Mars, in a certain JPL press release, was a fraud. To me, personally, the "face" has always seemed to be an accident of nature. But when I encountered this controversy a few days ago, I saw it as an opportunity for a fascinating mathematical challenge: an attempt to determine the exact algorithm that was used to "flatten" the landform in the image. I tried some Fourier tricks, and succeeded. The algorithm is a very simple 1x31 convolution, which resembles a crude and grossly exaggerated unsharp mask. It produces all of the expected effects: it eliminates vertical streaks, and destroys real 3D shading cues, and creates false shading cues, and gives the surrounding flat land a "grainy" texture, and creates "false shadows" above and below a small bright feature near the east edge of the image.
I experimented with similar convolution kernels in varying sizes, and it does indeed seem that the size 1x31 is particularly well suited to the purpose of deceiving the viewer about the true 3D shape of this particular hill.
Was the deception deliberate, or accidental? I don't know.
If the people who issued the press release were fooled by their own deceptive image processing, and really believed that the "face" was not actually a hill, as it appeared in the Viking images, but instead was a jumble of boulders encircled by a trough, as it appeared in their press release image, then their mistake was an honest mistake. But even if they were honestly ignorant of the problem, they could not have remained ignorant for long. Tim J. Parker's enhancement was released that same day, and anyone who looked at both images could see that the processing of the original image had been deceptive-- so deceptive, in fact, that it could reasonably be expected to inspire conspiracy theories.
The right thing to do, at that point, would have been to issue another press release, retracting the bogus image, and apologizing for the mistake. Someone decided not to do the right thing. This, then, would be two individual acts of Bad Astronomy: first, the original, possibly accidental, use of bogus data in a press release; second, the decision to refrain from issuing an apology and retraction, when they learned that an image in their press release was bogus.
If this story has a moral, then I suppose it must be something like this: when you attack Bad Astronomy, be conscientious, and fight fair. If you cheat, then you may get caught, which will only make the situation worse. If you innocently use bogus data by accident, then you'll have to issue an apology and retraction, which will only make the situation worse. If you accidentally use bogus data, but then you deliberately refuse to issue an apology and retraction, then... well, then we're back to cheating; see above.
--------
Update, 1/15/2008: I've uploaded the full-size animation, with no cropping or scaling, to an image host which doesn't have a file size limit:
(catbconv.gif, 512x480 pixels, 12 frames, 3,193,898 bytes, sha1sum 6b4d1f23915d3342e5888363d409fdaa97b82f17)
http://s2.supload.com/free/catbconv-...3559.gif/view/
--------
The text of this message is released into the public domain.
Attachment: an animation which shows the convolution in action.