Interesting read. Thanks for the write up.
 

This phenomenon can also be observed when training multiple probes with different random seeds: they converge to distinct directions, again showing lower cosine similarity than expected.

do you know why this happens? shouldn't the different seeds be converging to the same effiecient directions, like you say here : 

while DIM provides one direction within the refusal cone, probe-based methods converge toward a different, more efficient direction, essentially sampling distinct regions of the same cone

relevant article I just read - On the origins of empathy for other species

Among the Mbuti foragers of West Africa, ethnographer Colin Turnbull noted that many magical rituals invoking animal spirits ‘convey to the hunter the senses of the animal so that he will be able to deceive the animal as well as foresee his movements’ (emphasis mine). Such an ability to take on the thoughts of others may have similarly scaffolded an understanding of one’s self

In many of these cases of deception and mimicry, we can find explicit references to having to think like an animal thinks, act like an animal acts, and pretend to be an animal. These forms of mimicry include not only mimicking the behavior of prey animals, but also mimicking the behavior of their predators in order to become better hunters. 

...  By allowing the thoughts of other organisms into our minds, we built a pathway to developing empathy with our prey. Like the rest of the world’s predators, humans have to kill their prey to eat it. Unlike the rest of the world’s predators, we have the capacity to empathize with our prey. 

Outside of a Western context, the ethnographic record is rife with explicit examples of hunter’s guilt, or the pain felt from taking an animal’s life. I opened this piece with one example of a trapper’s guilt from Willerslev’s book Soul Hunters, but numerous examples from his ethnographic experience are recorded, such as in an interview with one young hunter who stated, ‘When killing an elk or a bear, I sometimes feel I’ve killed someone human. But one must banish such thoughts or one would go mad from shame.’

Examples like these are abundant. Consider this excerpt from Robert Knox Dentan’s book on the Semai of Malaysia, where one hunter discusses the guilt felt after his hunts: ‘You have to deceive and trap your food, but you know that it is a bad thing to do, and you don’t want to do it.’ Dentan goes on to state, ‘trappers should take ritual precautions like those associated with the srngloo’ (hunter’s violence) . . . Animals, remember, are people “in their own dimension,” so that the trapping he is engaged in is profoundly antisocial, both violent and duplicitous.’

"the tradeoff of pushing the complexity into the representation or the traversal of an algorithm"
could you say more?

Now, as people have pointed out, it is tricky to use split-brain cases to draw conclusions about people who have not had their hemispheric connections severed. But this case looks very similar to what meditative insights suggest, namely that the brain constructs an ongoing story of there being a single decision-maker

choice blindness experiments are a great example of this. people seem to change their actual preferences after having to come up with a justification for something they didn't even prefer but thought they did

in that attentional blink experiment why is T2 visible ≈75% of the time at first when lag is 100ms ? I seem to be missing something because it contradicts the finding