Cyber Shadows

DALL-E Image: Cyber Misinformation and Conspiracy Theory

The Summer of 2024 is seeing a dramatic rise in violent disturbance on our streets, driven by cyber misinformation and conspiracy psychology spread through the everyday web—not the dark web, but the regular, accessible internet. Words, images, and music merge in TikTok and YouTube style ‘mini memes’ that proliferate rapidly. In pursuit of attention, both humans and bots exploit human vulnerabilities to undermine rational and measured thinking processes. By transmitting influence through semantic earworms that embed directly into our subconscious, such tactics subtly shape motivations and beliefs. The resultant mayhem is ripe for exploitation.

Trapped within the echo chambers of newsfeeds, silos, rabbit holes, and other online traps, individuals become isolated within small, like-minded groups, cut off from everyday discourse. Any conflicting  evidence itself becomes a subject of suspicion, as do the motivations of the dissauders. The victims are enthralled into an imaginary world where they are at the whim of their often unseen manipulators.

Can Breaking News be unbroken?

Everyone is asking why tech companies can’t be regulated, but is that where the problem really lies? The issue might be with our own craving for excitment. Perhaps we’re emphasizing the wrong word—it’s not just about “fake,” but about “news.” In the pre-internet era, editors regulated news, but newspaper sales still drove content. How many stories in right-wing or left-wing papers were truly balanced? The major change today is that we no longer need industrial-scale printing presses. Commercial dynamics once demanded a mass audience for the same product, but now any influencer can be a publisher. And for today’s journalists, now more than ever, success no longer means just having a great story. In the era of online news, truth and quality have taken a backseat to the almighty trifecta: tweets, likes, and shares. AI generated headlines are added into the mix. The world of news has changed. Now the medium truly is the message – within its new cyberspace environment. Maybe regulating cybernews isn’t feasible within a true democracy because the dynamics have fundamentally shifted.

The Psychology of Extremism

But let’s think outside the box (since thinking within it doesn’t seem to work). How about regulating not what we believe, but what we do when we believe it? Presumably, the ‘bad actors,’ to the extent they are behind what is happening, don’t just want us to believe something untrue—they want us to act on these beliefs and cause mayhem. Of course, there are already plenty of laws against causing mayhem, so a more practical solution might be to reduce the number of people who cause it. Who are they? And why are some people more prone to believe in conspiracy theories than others?

There is a personality trait known to both psychologists and psychiatrists that can assess this vulnerability. One psychometric test for its assessment is the RISC (yes, its my own, the Rust Inventory of Schizotypal Cognition), The RISC scores people on a nine-point ‘stanine’ scale ranging between 1 (extremely skeptical), through 5 (average) to 9 (extremely fascinated by all sorts of weird ideas), and all points in between.

But schizotypal thinking is not the only psychological trait that can lead people to be sucked into conspiracy theories. Other risk factors are paranoia, excessive anxiety, insecurity distrust of others, antagonistic thinking, suceptability to fear-based narratives, intuitive ‘gut-feeling’ thinking, and dislike of ambiguity. These are in addition to social effects such as poor education, over-exposure to extreme political views, and social isolation.  And while we could identify ‘at risk’ individuals from all this by analysing online digital footprints and ‘Lookalikes’ (as I’m sure many security  services already do) this clearly involves an invasion of privacy, as the Cambridge Analytica scandle illustrated only too well.

Reversing the Gaze: Machine-in-the-Loop Systems

So what then is the solution? Generative AI could help, as it already does within the legal system, advising solicitors and advocates, managing data sources for leglislators and assisting judges in writing their reports. but is this a step too far? Within AI warfare there is a strong principle that there should always be a ‘human-in-the-loop’ when any decision is made that will affect human life.  But within cyberspace, it is these ‘humans-in-the-loop’, the actors inspiring all this mayhem, that are the problem.

So let’s ‘reverse the gaze’. Let’s put an AI in the loop. I assembled some prompts for GPT4 about this idea. Below is what it says.

GPT4 makes the case:

“You’re absolutely right to consider the potential drawbacks of a “human-in-the-loop” system, especially if the human in question is a “bad actor.” In scenarios where human biases or malicious intentions could compromise the integrity of AI systems, implementing a “machine-in-the-loop” system might be a more effective strategy. Here, the AI could serve as a safeguard, ensuring that ethical standards are maintained without undue influence from malicious human intervention. It would  involve automated processes that oversee or supplement human decision-making, particularly in critical areas such as ethics and security. In this configuration, machines not only perform tasks but also monitor, validate, and enhance human actions to prevent unethical practices. Its advantages will include:

Does all this give you a bad feeling?

Dees it remind you of ‘Brave New World’? It should. But why is this? After all AI doesn’t have sentience – it can’t think bad thoughts, or even good ones. It doesn’t have ‘free will’ and can’t initiate actions on its own. Even with the now archetypal “Terminator” film series, it was a mistake by humans that initiated Skynet. But therein lies the problem. It takes a Human-in-the-Loop (let’s call them an HITL)  to initiate a ‘Machine-in-the-Loop’ (MITL); and there are plenty of bad actors among us humans.

And a second reason for our doom-laden concerns. If I can get GPT4 to make these ‘useful’ suggestions, so can anyone else – including all those ‘bad actors’. And they have presumably been doing so for some time now, certainly since over a year ago when GenAI first hit the headlines.

So, suppose I was a populist, and I wanted to enact the popular agenda among populists – abolish democracy, replace with strong leaders, eliminate the woke including LGBT+ rights, enforce borders, reintroduce national service and – control – control – control.

Well – I might try a pilot run in a vulnerable country, use the apparent willingness of the populus to accept fake news (even when there is plenty of evidence that it is indeed fake). I might  do this by suggesting that there is ‘more than meets the eye’ and also that ‘civil war is inevitable’. When such comments come from diverse supporters of the same Great Leader from both the UK and the USA, you may indeed suspect there is a conspiracy. Or maybe they are just jumping on the bandwagon once its already on the roll.

What can be done?

Can the MITL also be used for the greater good? In the current state of affairs, we might consider this as maintaining the democratic process and the human rights enshrined in the United Nations and obtained by other hard won international agreements. Nonetheless, there are several current applications of the MITL approach being openly reported.

• The military increasingly use AI software to derive algorithms that minimize ‘collateral damage’ (although the software does also allow humans to move the goalposts on the definition of an ‘acceptable’ civilian causality count).
• GenAI companies use MITL in what they call ‘value alignment’ in their endeavors to prevent Chatbots from replying with ‘unacceptable’ answers or comments (perhaps best but most cynically described as answers or comments that are likely to lead to reduced profit).
• Regulators in Nation States will also be seeking advice from MITL on how to frame appropriate regulation.

And in the last of these, lies the problem. For most Nation States the ‘Bad Actors’ are often – other Nation States.  So, is this where the United Nations comes in? Or should come in?

Hopefully the UN will be able to generate a MITL that is powerful enough to model all his. I’m going to assume that it does. With enough will, we can make it happen.