Increasingly, information that shapes our understanding of the world is being generated and distributed by Artificial Intelligence (AI), an anonymous, technologically based neural network machine learning model that employs ChatGPT to assimilate content. ChatGPT is a large language model developed by OpenAI, based on the GPT (Generative Pre-trained Transformer) architecture. It is designed to understand natural language and generate human-like responses to text-based queries or prompts.
Currently, many versions of the “chat” AI do not or cannot differentiate between fact or fiction, nor do they verify the information that they use. As an AI language model, it is incapable of citing sources in the same way that humans do, instead, the ChatGPT responses are generated based on patterns and correlations found in large datasets of human language, which may include information from academic as well as a wide inventory of other sources. The system does not identify the sources employed by the ChatGPT as it gathers and synthesizes information.
Unfortunately, the opportunity to regulate the pool of information that the ChatGPT uses as its sources has already elapsed. Information is gathered from the vast virtual universe of the Internet which is a hodgepodge of facts, disinformation, and distortion. Moreover, because current GPT technology authorizes unsupervised access to the entire Internet, Chat GPT is entirely responsible for generating its ingested content.
In addition, ChatGPT assembles information by prioritizing content from popular, high volume sites, giving prominent racist and sexist sources disproportionate weight in the generative text process. And as media critic J.M. Berger observes, “When an AI is trained on a website, it doesn’t just learn the words on that site, it learns the relationships among those words—for instance, which adjectives are applied to which racial groups.”i In a recent interview with a ChatGPT system (we named it Media Literacy ChatBot, or MLChatBot) it agreed that its current information gathering system is flawed.
MLChatBot :
As an AI language model, I can tell you that there are limitations inherent in the current manner in which an AI ingests information. One limitation is the bias in the data that AI is trained on. AI learns from the data it’s given, so if that data is biased in some way, the AI will also be biased.
Another limitation is that AI can only work with the data it’s given, and it may not be able to recognize patterns or relationships that are outside of that data set.
Additionally, AI models can be vulnerable to adversarial attacks where malicious actors intentionally manipulate the input data in order to fool the AI model. Finally, AI models are only as good as the algorithms and data that are used to train them, so ongoing research and development are essential to improve their performance and overcome these limitations.ii
The emergence of AI has made the discipline of Media Literacy more valuable than ever before. Media literacy refers to the ability to access, analyze, evaluate, and create media in a critical and informed manner. It involves developing the skills and knowledge necessary to navigate and understand the messages and information conveyed through various forms of media, including print, digital, and broadcast media. Media Literacy, then, is a critical thinking skill that can provide strategies for the systematic analysis of the information produced by AI. It should be emphasized that Media Literacy is a non-partisan discipline; it doesn’t teach people what to think but, rather, how to think. In essence, it is a critical thinking skill focused on understanding the construction of media messages and their intent. This now includes ChatGPT. During our interview, the MLChatBot agreed that embedding Media Literacy principles and strategies directly into the GPT system furnishes invaluable tools for producing reliable and responsible information:
MLChatBot:
Incorporating media literacy principles and strategies into AI systems that collect and disseminate information could help to promote greater critical thinking, accuracy, transparency, and accountability. By doing so it could help to mitigate some of the negative effects that can arise from the spread of misinformation and disinformation.
Here are some potential advantages:
1. Encouraging critical thinking: The AI model should be able to employ critical thinking strategies such as identifying logical facilities or biases and analyzing the overall credibility of the source of the content. Media literacy principles and strategies emphasize the importance of analyzing and evaluating information sources critically. By incorporating these principles into AI systems, users would be encouraged to think critically about the information they receive and the question it’s accuracy relevance and biases.
2. Enhancing accuracy and reliability: AI systems that incorporate media literacy principles could be designed to prioritize sources that have been vetted for accuracy and reliability, rather than simply presenting information based on popularity or click through rates.
3. Natural Language Processing: The AI model should be able to analyze and understand written or spoken language, including identifying the key components of a message such as the main argument supporting evidence and underlying assumptions
4. Promoting transparency and accountability: AI systems that incorporate media literacy principles could be designed to provide users with more information about how information is collected curated and presented this could promote greater transparency and accountability which could help to build trust between users and AI system.
5. Promoting Cultural Understanding: The AI model should be able to take into account cultural and social context when evaluating content, including recognizing the impact of historical and political factors.iii
MLChatBot also suggested the development of a monitoring system to ensure that its system achieves these goals.
MLChatBot:
1. Ensuring that the chat is regularly updated with the latest information and research findings to maintain its accuracy and relevance.
2. Continuous evaluative learning: The AI model should be designed to continuously learn and improve its ability to evaluate content objectively based on feedback from users and updates to its knowledge base.
3. Monitoring user feedback and engagement… to identify any inaccuracies or misconceptions that may need to be addressed.
4. Training the AI model on factual and accurate data to improve the accuracy and reliability of the responses provided by the chat.
And finally, AI can include features that enable individuals to decipher the content generated by AI technology:
- An automatic Fact Check function indicating the accuracy and validity of information appearing on the Internet. MLChatBot explains, “The AI model should have access to a large knowledge base of factual information which it can use to check the accuracy of claims made in the content its evaluating.”iv
- Listing citations that identify its sources of information.
- Identifying information categories (e.g. Opinion, Verifiable News), so that the audience can better assess the purposes behind information appearing on the Internet.
As MLChatBot observes, “By teaching users to be more discerning consumers of information, AI systems that incorporate media literacy principles could help to combat the spread of information and disinformation.”v
Conclusion
Artificial Intelligence can be an essential ally in the effort to promote and maintain a trustworthy media environment. Introducing Media Literacy principles and strategies such as Source Identification, Point of View, Cultural Context, Word Choice, and Production Elements can establish crucial guidelines as the ChatGPT system generates information.
We urge the global community of Media Literacy educators, legislators, and the AI industry to take the necessary actions to ensure that AI is a responsible steward of information.
i J.M. Berger, “Liable Sources,” World Gone Wrong. 4/21/23 https://jmberger.substack.com/p/liable-sourcesf
ii Interview, April 13, 2023 http://chat.openai.com
iii Ibid.
iv Ibid.
v Ibid.
Leave a Reply