In last week’s post, I focused on an AI-generated image from the Paris protests that went viral across social media. The fake image, supposedly showing an elderly couple being robbed and humiliated in their underwear, was quickly disseminated by diverse actors, all hoping to inflame tensions in France. Each group used the image to cement its narrative of events. Right-wing politicians, journalists, pundits and influencers claimed the image showed Muslim immigrants cruelly abusing an elderly couple. Left-wing pundits claimed the image showed French fascists abusing an elderly couple amidst the chaos of mass protests. Others claimed that the robbers were left-wing anarchists, or young thugs that had been failed by a deteriorating education system or even pro-Europe protesters that love Europe but hate France.
I asserted that the AI image of the elderly French couple was demonstrative of ‘Generative Outrage’, or the deliberate production and circulation of AI-generated visual content designed to provoke public anger, moral indignation, and political hostility by depicting fabricated or manipulated events. Unlike traditional disinformation, which seeks to convince publics that a particular claim is true, Generative Outrage transforms publics into witnesses of crimes, horrors, and atrocities. Through these images, social media users are transformed into witnesses of events that never occurred, and as witnesses, they have a moral obligation to cry out against these fabricated events. It is the transformation of publics into witnesses that distinguishes ‘Generative Outrage’ from other forms of deepfakes and AI-generated images.
Moreover, I asserted that the power of images lies in the fact that they cannot be unseen. Images that generate a strong emotional response become seared into one’s memory; they will forever be recalled as they have been tied to an emotional response. The important question soon becomes how can we combat the power of such persuasive AI-generated forms of disinformation? The answer, I argue, lies in changing the emotion that the visual elicits. Unlike traditional fact-checking, which operates primarily through a cognitive mechanism of correction or debunking, I envision a form of affective reframing where the same AI image is used but is tied to other emotions such as humor. Affective reframing involves deliberately altering the emotional meaning of a visual through its reuse in a different context.
For example, the fake elderly French couple can be transported and re-contextualized in such a way that elicits a humorous response. In the fake AI image, viewers interpret the elderly couple as helpless victims of a brutal assault. Their appearance becomes associated with vulnerability, humiliation, fear, and anger. But one can re-contextualize the couple and create an amusing new context. For example, by situating the elderly couple in a school and publishing a series of POV reels (point of view). In one reel, the elderly couple can show up half-naked to pick up their grandson from school with the caption “POV: When it’s grandpa’s day to pick me up from school”. This might generate a different emotional response from anger, be it mirth, amusement or familiarity.
A second reel can place the elderly couple in their underwear in a supermarket, saying “POV: Your grandparents just heard that butter is 50% off”. A third can situate both elders in a busy daytime street, saying “POV: Your grandparents forgot it’s daylight saving time”. A fourth can place the couple in a workplace environment, saying “POV: It’s bring-your-grandparents-to-work day”. The goal here is not to change people’s memory nor to erase the original fake image but to change the associations attached to it. So, instead of being remembered exclusively as victims of a supposed atrocity, the couple might also be remembered as the amusing grandparents from a viral reel or meme.
Studies partially support this approach, with scholars finding that misinformation can continue influencing people’s reasoning after correction or debunking. Second, scholars have found that humor could act as an emotional antidote to misinformation. These studies suggest that humor can attract attention and shape how people process corrective information. Moreover, an eye-tracking experiment comparing humorous and non-humorous misinformation corrections found that humor increased attention to corrective imagery.
The ultimate goal of using humorous images to combat the power of generative AI images is narrative displacement or the creation of a competing interpretation of a visual that may weaken the dominance of its original narrative. In other words, the elderly French couple is deliberately displaced from a cruelty narrative into a humorous narrative, changing not the memory of an image but the emotions and associations it elicits.
It is impossible to make people unsee an image, especially one that elicits a strong emotional response. But perhaps the goal of combating disinformation can be to alter the emotional response people have when recollecting an image, such as substituting rage with mirth. The pitfall of this approach is that one may soon be accused of callousness and cruelty as social media users are horrified by seeing the robbed elderly couple transformed into a viral video. Hence, this approach of narrative displacement necessitates corrective information in the form of “POV: Those fake AI grandparents forgot it’s daylight saving time”. Or “POV: It’s bring-your-fake-AI-grandparents-to-work day”. The objective is not to ridicule victims of violence, but to undermine the emotional power of fabricated victimhood.
This post suggests that in the age of Generative AI, in which fake AI images can stir powerful emotions and even deepen social rifts and political polarization, new approaches may be required to combat disinformation. This post argues for fighting visuals with visuals through two processes. The first, affective reframing, is the process through which an image becomes associated with new emotions by being placed in a different interpretive context. This leads to the intended outcome of narrative displacement, whereby a new interpretation competes with, and potentially weakens, the original interpretation and context of an image. This could be a way of quickly and effectively combating visual forms of disinfromation.
Leave a Reply