What to do when we can’t trust our own eyes (or at least, the videos we are looking at.
I spoke last weekend at a panel discussion on Navigating Lies, Deepfakes & Fake News, organised by McPherson Independent. This a group promoting the idea of an independent community candidate in the electorate of McPherson south of Brisbane, currently held by the (centre-right to right) Liberal National Party. It’s part of the broader disillusionment with the two-party system we are seeing in Australia and also in the recent UK election.
It was a great discussion. I prepared some preliminary notes, which I’ve provided below. Comments and constructive criticism most welcome
Lies, Deepfakes and Fake News
It’s important to understand that there is nothing fundamentally new here. Both propaganda and forgery have been around at least since the invention of writing.
Deepfakes raise two issues, because they are more realistic and potentially more convincing than ever before.
First, their use in harassment, particularly sexual harassment, is more problematic and distressing by virtue of their greater realism. A range of legal and social responses are needed, but this is outside my area of competence.
Second, a form of evidence we have assumed to be reliable (video) can now be faked. This has happened before with forged paper documents, and then with photography. While there may be technical solutions, the main response must be social and relies on trust, in the form of assured provenance. If we know that a photo or video was taken by someone we trust and transmitted to us through a trustworthy process we can believe it to be accurate.
Similarly, there is nothing new about misinformation. In particular, panics about social media fail to take account of the longstanding role of traditional media (notably, but not exclusively, the Murdoch press).
Given the reluctance of mainstream journalists to attack each other (they might end up working for the same organisation, after all) social media outlets provide an opportunity for critical comment. This comment in turn is bitterly resented and misrepresented by journalists, who amplify the most offensive examples (sexist attacks on women, for example) to justify treating all their critics as “trolls”.
Misinformation works at both an individual and a social level. Individually, we can’t check everything and are prone to ‘confirmation bias’, paying more attention to things that confirm our existing beliefs. Again, we need to find trustworthy sources and (equally importantly) dismiss sources that have been shown to be untrustworthy. As regards confirmation bias, the heuristic “if it sounds too good to be true, it probably is”.
The bigger problem is at the social level. Social groups can help to inform their members, or they can promote confirmation bias, to the point that their members actively desire misinformation. This was true at one time of some groups on the political left and is now true of the political right as a whole, most notably in the US.
There is no effective strategy to correct misinformation once it is firmly established within a cohesive social group. The task is comparable to trying to convert a religious group to an alternative religion or to non-belief.
The only real response is to focus on loosely attached members of misinformation groups, and seek to point out how they have been misled. This is a very slow process, but there are plenty of examples of success.
Climate change provides a good example. Attempts to convince rightwing denialists using a variety of strategies (factual evidence, clever framing etc) have gone nowhere. But, over time, everyone open to being convinced has come to accept the reality of human-caused climate change. Those promoting misinformation have never admitted error, but have been forced to change tack, shifting from science denial to attacks on clean energy and promotion of nuclear power..
{ 17 comments }
Cervantes 07.14.24 at 2:59 pm
Nothing new of course, from the Donation of Constantine to the Protocols of the Elders of Zion, blood libels, witches, and before that, of course, Exodus and the Book of Joshua (none of which actually happened). What is disturbing to us I think is that we actually did live more in a realm of True Facts for a while, with some sway for journalistic ethics, relatively trustworthy recording devices for sound and pictures, and the impressive achievements of science-based technology. Of course there was always a minority of people with nutty beliefs. What’s disturbing is that we seem to be losing what progress we had made. There are plenty of reasons for that, but Rupert Murdoch has done more than his share, along with Mark Zuckerberg, who introduced a form of technology that allows bad actors to amplify disinformation at little cost. I don’t have a prescription.
Peter Dorman 07.14.24 at 6:45 pm
I agree with most of what this post says, especially about the long history of disinformation. But trust, while part of the answer, is not all of it. First, we need to criminalize digital impersonation, even at the cost of infringing on legitimate satire. It’s been proposed (somewhere) that it be illegal to generate and publish any AI-produced text using the first person: take the “I” out of AI. Second, we have a problem with algorithms established to monetize attention — they also create a network infrastructure for cognitive bubbles. Perhaps a combination of regulation and public funding can lead to the use of other search and push alg0s that are less conducive to serving as disinformation conduits.
I’m pretty confident in my first suggestion, less in the second. I know a lot of very smart minds have been puzzling over this, and I’m interested in what ideas people have come up with.
Tm 07.14.24 at 9:23 pm
What stands out to me is the absolute failure of the mainstream legacy media to deal with the disinformation epidemic. Most obvious in the case of the US where Trump‘s constant lies have mostly been amplified instead of corrected by the so-called reputable news media. And I see very similar tendencies for example in Germany, despite the existence of a system of publicly funded media that were supposed to counter fake news with fact checking and well researched journalism. (Not to mention pure propaganda outlets controlled by fascist billionaires whether it’s Murdoch or Springer or Bolloré).
Now it’s clear that the legacy media cannot solve the disinformation problem alone, now that we have this the huge uncontrollable social media ecosystem. But they could at least not have made it worse. And that’s what they did and what they consistently do. And this is a big part of the problem because the participation of the reputable news media plays an important role in legitimizing disinformation.
Alex SL 07.14.24 at 10:07 pm
Apart from relying on trusted sources, another solution would be to have a decent grasp on what is plausible and what is implausible. Many deepfake videos of politicians seem to fall into the categories “why would any real person ever behave like that” or “this person would never say these words, as they stand in diametrical opposition to their character and beliefs”. Sadly, as everything from Pizzagate to Covid denialism amply demonstrates, most people do not have a good feeling for what is plausible.
Regarding the beginning of the post, I understand disenchantment with the two large parties. I do not understand why so many people are so enamoured with the concept of independents. Surely they must realise that there is an enormous advantage to organising as a party that can share resources and coordinate strategy? Are they unaware that parties other than the two large ones already exist, e.g., the Greens?
KT2 07.15.24 at 5:16 am
We need some very serious K-12 education on “Lies, Deepfakes and Fake News”.
And cognitive science.
We are all now, personally, able to have bespoke tailored messages – microtargeting – delivered to your screen. At an appropriate time. As seen by “some part of your brain is overriding the plain truth about the path of the object”… which… “You can’t seem to consciously override the ‘wrong’ interpretation.”.**
JQ: “What to do when we can’t trust our own eyes (or at least, the videos we are looking at.”
Or brain.
** “Reality” is constructed by your brain. Here’s what that means, and why it matters.
“Fix your gaze on the black dot on the left side of this image. But wait! Finish reading this paragraph first. As you gaze at the left dot, try to answer this question: In what direction is the object on the right moving? Is it drifting diagonally, or is it moving up and down?
“Remember, focus on the dot on the left.”
[Visual illusion graphic]
…
[2nd Visual non illusion graphic]
…
“But you don’t need an fMRI to conclude that some part of your brain is overriding the plain truth about the path of the object. You can see it for yourself. “The remarkable thing is that — even when you are told what is happening — you still see it in the illusory form,” Justin Gardner, a Stanford University neuroscientist who wasn’t involved in this study, said in an email. “You can’t seem to consciously override the ‘wrong’ interpretation.”
…
https://neuroscience.stanford.edu/news/reality-constructed-your-brain-here-s-what-means-and-why-it-matters
Damn. I couldn’t “…seem to consciously override the ‘wrong’ interpretation.”!
Solution. We all have to have;
– Our data – life – secured on our device
– Develop our own personal AGI with sensors, watchers, and an immune system.
– have our own models and weighings with some oversight by social mores and government.
Abive is potentially in the ‘lab’ by 2030. For example googl already has an AI running in an AI. Therefore we may be able to run out own AI in their AI. Karl Friston is at Verses ai attempting AGI and theory of mind models.
Delivered to every person? After we manage climate change. Or get one world government And tangerine tyrants consigned to the dustbin of history.
c1ue 07.15.24 at 3:17 pm
Concern over deepfakes is laughable given that all of the Western governments are deep into the “disinformation censorship complex” rathole, which in reality is little more than censorship of dissent and free speech.
John Q 07.15.24 at 8:23 pm
Alex SL @4
This is very Australia-specific.
The Greens are mostly competitive in left leaning seats which would normally be won by Labor. What has emerged recently is a national movement to promote centrist independents in seats like McPherson, where neither Labor nor Greens has any chance. tThey have been called “teals”, that is blue-green. This movement has some, but not all, of the characteristics of a political party.
Alex SL 07.15.24 at 9:56 pm
John Q,
I am an Australian citizen, and I know about the Teals, I just find it silly that these people try to pretend that they aren’t building what will ultimately have to become a political party if it is ever meant to actually be functional as a political force. Partly my take may also be informed by the “independents” of 1990s-2000s Germany, where I grew up; they were (probably still are?) party-ish conservatives who competed at local elections and pretended they didn’t have the exact same values as the conservative CDU but nearly always formed a coalition with the CDU if neither got the majority.
So, to me, the charitable take on independent is conservative who genuinely doesn’t understand why parties exist (like gold bugs who don’t understand why deflation is a problem, or certain billionaires who hate public transport and then re-invent a less efficient version of it), and the less charitable take is conservative who tries to trick voters into thinking they aren’t just another fiscal conservative, although in fairness I think the Teals are at least genuine about trying to be better on climate.
engels 07.15.24 at 10:18 pm
Tangential but I find it strange that screenshots are now conventionally accepted as proof of deleted tweets, posts, messages, purchases, etc. Umm any moron could fake that, without using AI.
MisterMr 07.15.24 at 10:37 pm
I’ll just note that relying on trusted sources or on what feels believable is exactly what makes conspiracy theories or fake news possibile, as wevtend to trust and accept as realistic those informations and/or people who generally agree with our previous ideas.
Tm 07.16.24 at 1:16 pm
Alex: you are referring to the “Freien Wähler” (“free voters”). It is indeed an oxymoronic name for a political party. But these groups used to be restricted to local networks and were not supposed to be political parties, and yet they have meanwhile formed state parties and are competing in state and European elections as just ordinary right wing parties. Outside of Bavaria they are mostly irrelevant but in Bavaria they have eaten into the CSU vote and become their junior governing partner. That the Bavarian party leader has a Nazi past (he was caught with a holocaust denial flyer when he was 16 or 17) has increased his electoral appeal.
John C Quiggin 07.16.24 at 9:51 pm
@MisterMr There is indeed a problem of “epistemic closure” (a misapplication of the term)
The only cure here is to pay attention to what happens when your putatively reliable source makes claims that are directly falsifiable, such as “global warming isn’t happening”. If they acknowledge error and correct, that’s good. If they don’t, you need to start checking them more closely.
J-D 07.17.24 at 9:50 am
I agree with John Quiggin’s description of the teal independents as having ‘… some, but not all, of the characteristics of a political party …’
I would also agree with the suggestion that not having all the characteristics of a political party is a limit on the political influence they can have. However, I wouldn’t agree that not having all the characteristics of a political party prevents them from having any political influence. They already has some political influence, although not as much as a successful political party could have. What is hard to calculate is whether they would be successful if they tried to become a political party, with all and not just some of what they entails. One possibility is that the attempt would make them a more potent political force, but another possibility is that it would destroy the movement. If they decide to make the attempt (which I suppose they may) then we will find out; if they decide not to, then we’ll never know for sure.
SusanC 07.21.24 at 11:32 am
Arthur Conon Doyle was apparently taken in by the Cottingley Fairies.
We have been through a historically fairly brief period of time when people believed in photographic evidence. It’s coming to an end now, and always was somewhat questionable. (See the aforementioned fairies; also Stalin without Trotsky)
somebody who remembers that iraq "sent their wmds to syria" 07.21.24 at 11:23 pm
the cottingley fairies incident is highly instructive here. it shows that you can’t educate your way out of this situation. believing in the material reality of spirits and ghosts was very common at the time and so doyle was prepared to believe it, and did. he believed in the photos because he believed in spirits and ghosts, not the other way around.
similarly, around sixty or seventy million americans believe that hillary clinton has been replaced with a clone because trump threw the original into guantanamo bay and had her hanged for participating in a pizza-based pedophilia ring and trying to overthrow america with her e-mails. the people who believe this don’t believe it because of a bunch of 8chan posts, or because they saw a picture with a giant red arrow on it. rather, they believe “all American democrats support pedophilia, rape and infanticide because they’re satanist communists”, and THEN they subscribe to the 730,000 person Facebook group “The Real Truth The Marxist NYT Won’t Tell You” to see the picture with the red arrow on it. If there’s no such picture many will create the picture themselves, to express what they believe!
you cant educate your way out of this. for the most significant beliefs in their lives, people decide what they believe before they see the evidence, and they don’t change their minds.
KT2 07.22.24 at 11:43 pm
JQ’s backyard… “It seems deepfake content has reached the Queensland election.”
Live by the sword; “[Qld Premier] Miles is an avid TikTok user and regularly shares videos of himself”
Die by the sword…
“Queensland premier denounces AI-deepfake video posted by state opposition of him…” …
“Last week, the Liberal National Party posted an AI video of Steven Mileson their TikTok account, showing the premier dancing to ‘Closer’ by Ne-Yo.
“A caption over the video reads, “POV: my rent is up $60 a week, my power bill is up 20%, but the Premier made a sandwich on TikTok,” and the video has a disclaimer: “Creator has labelled as AI-generated.”
“Miles is an avid TikTok user and regularly shares videos of himself chatting to the camera about state government policy, while making a sandwich.
“In a statement to the ABC, Miles said that it was “appalling and disgusting” the state opposition leader, David Crisafulli, had “stooped to using AI and deep fake videos to attack me.”
“Mr Crisafulli needs to explain why he thinks this is OK. It’s a test of his character. In a time when misinformation is everywhere, we as politicians have a duty to communicate with our audiences and to voters clearly and honestly.
“Separately in May, the Australian Electoral Commission said it expected AI-generated misinformation at the next federal election, but warned that it doesn’t have the tools to detect or deter it. You can read more on this below:
(AEC warns it doesn’t have power to deter AI-generated political misinformation at next election
Read more)
https://www.theguardian.com/australia-news/live/2024/jul/23/australia-news-live-politics-julian-assange-kamala-harris-albanese-dutton-biden-us-election-labor-coalition-greens-weather-nsw-vic-qld-ntwnfb?CMP=share_btn_url&page=with%3Ablock-669ecaae8f083fc6a547bcc4#block-669ecaae8f083fc6a547bcc4
KT2 07.24.24 at 3:15 am
JQ you said; “There is no effective strategy to correct misinformation” … but not though want of trying…
“Mapping the field of misinformation correction and its effects: A review of four decades of research”
November 2021
Social Science Information 60(41):053901842110537
Authors:
Qinyu Ee
The University of Tokyo
Osamu Sakura
The University of Tokyo
Gefei Li
Citations (7)
References (67)
Abstract
“Why people still rely on misinformation after clear corrections is a major concern driving relevant research. Different fields, from psychology to marketing, have been seeking answers. Yet there remains no systematic review to integrate these theoretical and empirical insights.
“To fill the gap, this article reviewed 135 articles on misinformation correction and its effects written before 2020 to examine the knowledge generated in the field.
“Our findings indicate a consistent interest on this topic over the past four decades, and a sharp increase of relevant scholarly work in the last ten years.
“Nevertheless, most studies have been built upon psychological inquiries and quantitative methodologies. What is lacking includes longitudinal measurements of debunking effectiveness, theoretical insights beyond cognitive sciences, methodological contributions from qualitative approaches, and empirical evidence from non-western societies. With this analysis, we propose worthwhile focuses for future exploration.”
https://www.researchgate.net/publication/356168606_Mapping_the_field_of_misinformation_correction_and_its_effects_A_review_of_four_decades_of_research
Comments on this entry are closed.