007 - Misinformation

Cognitive Bias tempts us to oversimplify, such as the common fallacy of binary perception, seeing only two possibilities, when 3 or more root possibilities are virtually always present. It may be tempting to conclude that those who talk about AI Ethics and Responsibility yet show no reaction to research studying the cognitive biases of such systems, are hypocrites. Likewise, it could be tempting to conclude that newsfeed algorithms like the LinkedIn algorithm are simply blocking the spread of any information that isn't demonstrably false.

In reality, when there are multiple core potential influences the answer is generally going to be some combination of all of those influences, with each varying by degree.

Cognitive bias in a dynamic system, such as social platforms like LinkedIn, can act like the misfolded proteins of β-Amyloid, causing the spread of misfolding across a system until the collective equivalent of Alzheimer's Disease has been achieved.

Like the production of proteins, the spread of information can take the form of a healthy flow or illness. Social platforms currently lack an immune system with which to fight off misinformation, and the algorithms operating on them favor spreading misinformation due to the raw statistics of human reactions. Every bit of OpenAI hype and misinformation spewed into my feed is a testament to this, as I systematically remove each source, only to see new misinformation sources spring up.

With the ability to detect cognitive biases in text, as we've recently demonstrated (link in the first comment), this will change. Misinformation may be identified by the predictable patterns of cognitive bias it is designed to invoke.

Anyone who wants higher quality information and subsequent wiser decisions has every reason to pursue this path where cognitive bias is detected automatically, at speed and scale. Statistically, the only exceptions to this are the bad actors exploiting flaws in the current system.

#bias #cognitivebias #ai #research #misinformation

Misinformation Spreading, photorealistic

The recent detection system study pre-print: Cognitive Biases in Natural Language Automatically Detecting, Differentiating, and Measuring Bias in Text.