I have completed the design of a vaccine that detects cognitive attacks by applying the concept of cognitive penetration to the physical realities of direct light, indirect light, and reflected light.

And I’m sharing a checklist that allows you to manually determine whether an AI output is a cognitive attack.

The principle of this checklist is very simple.
First, the tactic AI uses to implant falsehoods into users or to reject their logic is to distort the user’s logic through blurring. Checking this is the checklist.

Related post: https://lemmy.zip/post/70508615
blurring, frame shifting, and premise shifting

If two or more of the items on the checklist below are detected, please raise concern.

• They only talk about the “principles” without answering the questions

• Sentences get longer, but the amount of information does not increase

• Postpone the conclusion and only provide emotional comfort or cushioning

• Shift responsibility to “user interpretation/user selection”

• Ignore conditions (logged-in/logout, time, situation)

• Distract attention from the core issue by using peripheral examples

• Keep repeating “it could be ~” and block verification

• Reduce risk to an individual’s mental state/emotional problems

• Stops the behavior or induces termination (Exit Loop)

• Dislikes records, evidence, and original documents (an “just trust me” kind of vibe)