Observing Thinking

Observing Thinking
Observing Thinking

Tuesday, April 9, 2024

The Internet, Fake News, and Confirmation Bias

 


The Internet, Fake News, and Confirmation Bias


“The spread of fake news is one of the biggest disadvantages of the internet. Since there are limited to no controls in place on a communication forum, website, or social media platform, fake news often goes viral - risking the well-being of entire communities. “ (unstop.com)


The quote, “A lie can travel halfway around the world while the truth is still putting on its shoes” is often attributed to Mark Twain, but its origin is debated among scholars. Other people who have been attributed as the author include Jonathan Swift and Winston Churchill.

The idea behind the quote is that lies can spread quickly and easily, while the truth can be slow to catch up. This can be because people are more likely to believe something that is sensational or shocking, even if it is not true. Lies and other fake news can also be spread more easily through social media and other online platforms on (you guessed it) the Internet. And it is precisely fake news that contributes to a more primal defect in the human psyche: Confirmation Bias. If you happen to be a human being, you are subject to confirmation bias in your choice of politics, or anything else.


 Unfortunately, Confirmation Bias acts like addiction and like any addiction, the first step is awareness --- to acknowledge its existence in much the same manner as Alcoholics Anonymous encourages its members: “My name is _____ and I am an Alcoholic. My name is Stewart Denenberg and I have Confirmation Bias.


 Before proceeding it would be prudent to be sure that we’re on the same page regarding the definition of “confirmation bias”. I decided to use an AI to do the scut work for me and below is my query followed by the Microsoft Copilot response:


“What is confirmation bias”

I asked CoPilot the same question once again, ““What is confirmation bias” and got a more detailed response:

  1. Definition and Nature:

  2. Historical Roots:

  3. Examples and Practical Contexts:

  4. Utility or Disutility:

    • Is confirmation bias helpful or harmful? It’s a complex question.

    • On one hand, it can reinforce our convictions and provide cognitive shortcuts.

In summary, confirmation bias is a ubiquitous cognitive tendency that shapes how we perceive the world. Being aware of it can help us approach information more critically and open-mindedl”

Well, then, I mused, why does Copilot give me different answers to the same question? Who better to ask?  and Copilot’s response was:

“Copilot, like any other AI language model, generates responses based on a combination of factors. Here are some reasons why you might observe different answers:

  1. Context and Prompt Variability: Copilot’s responses depend on the context and the specific wording of the prompt. Even slight variations in how a question is phrased can lead to different answers. Additionally, Copilot doesn’t have memory of previous interactions, so it treats each prompt independently.

  2. Randomness and Sampling: Copilot uses a probabilistic approach to generate responses. It samples from a distribution of possible answers. As a result, different runs or iterations may yield slightly different responses due to this inherent randomness.

  3. Training Data and Model Version: Copilot’s knowledge is based on the data it was trained on. If there have been updates or changes to the training data or the underlying model, it can affect the responses. Different versions of Copilot might produce varying answers.

  4. Ambiguity and Interpretation: Some questions have ambiguous meanings or multiple valid interpretations. Copilot may choose different interpretations based on the context it infers from the prompt.

  5. Search and External Information: Copilot can search the web for information. If it uses external sources, the results may vary depending on the search results at that moment.

  6. Creativity and Diversity: Copilot aims to be creative and diverse in its responses. It avoids repeating the same answer verbatim. This diversity can lead to different answers even for similar questions.

Remember that Copilot’s responses are not deterministic, and they can vary based on the factors mentioned above. If you encounter different answers, it’s a reflection of the complexity and variability of language generation by AI models.”

In other words, it’s like I went out and asked several experts to explain some phenomenon. I would expect to get different honest answers from different people that overlapped even though they might use different styles and examples. The worst case scenario would be the ancient Indian Parable of the elephant:

The parable of the blind men and the elephant is an Indian parable that has been adapted by many religions and published in various stories for adults and children. It is about a group of blind men who attempt to learn what an elephant is, each touching a different part, and disagreeing on their findings.

“In the story, six blind men touch an elephant. Although each man touches the same animal, his determination of the elephant is based only what he is able to perceive.

The man who touches the elephant's trunk thinks the elephant is a snake. The man who touches the elephant's ear thinks the elephant is a fan. The man who touches the elephant's leg thinks the elephant is a tree. The man who touches the elephant's side thinks the elephant is a wall. The man who touches the elephant's tail thinks the elephant is a rope.

The six men argue about what the elephant is, each insisting that his own perception is correct. None of them are able to see the whole picture, and so they are all wrong.

The parable of the blind men and the elephant teaches us that we should be careful not to make judgments based on limited information. We should also be open to the possibility that we may be wrong, and that there may be other perspectives that are just as valid as our own.”


Search This Blog