Att påverka utan att märkas – om EU:s reglering av AI-driven subliminal teknik

Sammanfattning: In the EU proposal for a regulation laying down harmonised rules on artificial intelligence (AIA), article 5.1.a forbids the use of AI systems deploying subliminal techniques. As the meaning of the term subliminal techniques lacks a definition in the proposal, the intended meaning of the phrase requires interpretation to clarify the scope of this prohibition. Using a method based on the legal reasoning of the European Court of Justice, the meaning of the expression subliminal techniques is interpreted based on arguments from a grammatical, contextual, teleological and historical standpoint. After reviewing EU documents concerning the proposal as well as the current versions of AIA, it becomes clear that the intended meaning of subliminal techniques is broader than the ordinary meaning of this phrase. Although subliminal stimuli are generally described as stimuli that are beyond human perception, the presence of a binary threshold of perception is not fully supported by brain imaging research. From the descriptions of subliminal techniques in documents regarding AIA, as well as in formulations present in different versions of the proposal, there is support for interpreting the legislators intent as adhering to this non-binary view of subliminal techniques. Instead of strictly focusing on a perceptual threshold the determination of whether an AI deployed technique is regarded as a subliminal technique focuses on whether the technique in some sense targets human perception, as well as the degree to which this technique infringes on human autonomy.