I've been thinking a bit more about RAW's filter model and how it relates to memebomb effectiveness. A memebomb is an aphorism; it is not a novel (which invites someone to willingly enter someone else's reality tunnel and look through their filters). Instead of being drawn as positive space (like a cathode ray tube -- which moves electron beams around to light up portions of currently dark phosphor -- or a painting -- which fills an empty canvas) it may be more suitable to consider a memebomb as a negative space (like an LCD -- which takes white light and subtracts the colours not in the intended image -- or a sculpture -- which is a chunk of rock with all the non-sculpture pieces taken out -- or a filter).
Someone encountering a memebomb unexpectedly will look at it through their habitual filter. You can perceive a memebomb as a secondary filter, then: an effective memebomb will highlight inconsistencies in an existing filter by removing the context that obscures them. Alternatively, you can consider one's habitual filter as a given: a person projects meaning only insomuch as his filter allows, and the text merely further restricts possible meanings (a kind of reverse Gantzfield effect: by adding sensation you restrict the domain of 'normal' hallucinations or projections by forcing them to coincide with the sensed world).
So, under this model, an effective memebomb will consist of a juxtaposition between several elements of an existing worldview. The 'target audience' should agree with the elements but be predisposed to disagree with the juxtaposition (i.e., the juxtaposition is locally consistent but conflicts with something else in the reality tunnel). Thus, the memebomb causes cognitive dissonance by laying bare a contradiction in already accepted beliefs without adding any new information for the target to accept. The target cannot therefore choose to reject the memebomb: he has already accepted it, but not considered it in isolation. The target is forced to consider and rationalize this juxtaposition (expanding the filter by adding a bag on the side full of twisty reasoning and justifications, which -- being the quickly-erected defense against an already successful attack -- can be hotwired by other ideas).
For a memebomb to be successful under this model, there are some things it cannot do:
1) it cannot consist primarily of new material, because the new material can be trivially rejected
2) it cannot consist of already-rationalized contradictions ("Can God create a burrito so hot even HE can't eat it" has been rationalized in various guises by the Jesuits and others for centuries, and is therefore neutered despite being a legitimate contradiction in the Abrahamic-Scholastic theological idea of God; it will work, briefly, on the rare poor soul who has never heard it)
3) it cannot be perceived as nonsensical (i.e., few people will project an emotionally intense meaning upon "Colourless green ideas sleep furiously" -- though this does not factor out all uses of language generated without intent, as memebombs can certainly be created accidentally)
4) it cannot be immediately seen as an attack, because it will be rejected and defended against before it is processed
I suspect that following the above rules will result in a slightly higher signal-to-noise ratio in memebombs, but that many effective memebombs will be filtered out. Indeed, there are clear exceptions to the above rules. My intent was to systematize the problems I saw in many of what I perceived as 'bad memebombs' exceedingly common on the SOMA. I am open to tweaks, suggestions, and (necessarily) flames about how I am shitting up the board with talk about RAW &c.