Intelligence Sourcing Basics: A Primer as "Generative AI Pollutes the Internet to Death"
Questions to ask now that we need to question everything we see online.
Image generated by DALL-E 2
I’ve read a few articles recently suggesting the proliferation of generative-AI tools means the internet is about to get absolutely flooded with synthetic content of questionable quality and reliability. I’m not just talking about funny images of the Pope in a swaggy puffer coat or computer-generated sci-fi story submissions, but also deepfake news broadcasts spinning disinformation and AI-generated audio of world leaders behaving badly.
The MIT Technology Review warns “we may be witnessing, in real time, the birth of a snowball of bullshit” created by generative-AI platforms that will force us to question at every turn “whether what we read online is written by a human or a machine.”
Experts at the Copenhagen Institute for Future Studies in February 2022 forecasted 99 percent to 99.9 percent of the internet's content will be AI-generated by 2025 to 2030 if platforms like chatGPT open for public use (which they now have), ultimately driving us to a point where “the internet would be completely unrecognizable.”
Fellow Substack author Alberto Romero over at The Algorithmic Bridge in October 2022 wrote in a post titled “Generative AI Could Pollute the Internet to Death” that eventually these tools “will overflow the Internet with low-quality content.”
In a haunting piece I keep coming back to by designer, developer, and anthropologist Maggie Appleton, she argues that what we’re witnessing is like “an explosion of noise that will drown out any signal.”
Generative AI is poised to “overflow the Internet with low-quality content,” creating “an explosion of noise that will drown out any signal.”
This is, objectively, problematic for lots of reasons.
But it’s particularly concerning for anyone whose profession involves diligently and rigorously parsing through information to help illuminate fact from fiction. Teachers. Students. Librarians. Journalists. Researchers. Scientists. Parents. The list goes on.
One way intelligence analysts might navigate the oncoming onslaught of AI content is to double down on a rigorous application of sourcing tradecraft - a set of mental and linguistic frameworks that exist to help intelligence writers and consumers better understand the quality of information involved in any piece analysis. And if there was ever a time to be transparent and diligent about information quality, this might be it.
There is no universal standard across public and private sectors about how to apply sourcing tradecraft. Or any intelligence tradecraft, really - a topic I’ll save for another day.
But I’d argue there are some key questions to ask when you come across a piece of information to help decide if and how to leverage it for intelligence analysis.12 3
If there was ever a time to be transparent and diligent about information quality, this might be it.
Find the source 🔦
Are you using diverse information streams to help answer your intelligence question? Remember, information sources exist on and offline - don’t build your collection plan entirely around one platform, aggregator, or method.
Are you conducting a gap analysis to understand what new sources might help you find your answer? You’re likely biased to use the information you already have access to, but what if the answer you need is hiding somewhere else?
Analyze the source 🔎
Does this source help answer my intelligence question? If not, move on.
What is the intent of the source? Is it motivated to provide biased, deceptive, or inaccurate information? What tools can you leverage to understand the source’s motivation, potential bias, or even its intentional deployment of dis/misinformation?
Beware the game of telephone - have you found the primary sources? Are you working with the origin of the information, a source with first-hand access to it, or someone who knows someone who overheard something and is now reporting it as truth? For example: a source shouldn’t be the news story covering the CEO’s statement, it should be the CEO’s statement itself.
Is the information accurate and complete? Are there glaring omissions in the information or parts of the story that appear to be missing? Does the information conflict with other sources out there?
Is the information outdated, developing, or likely to be overcome by rapidly moving current events? Things move fast, make sure the information you’re building your analysis around is the most current version.
Can you validate the information through additional sources? How solid do you feel about the veracity of this source and can you find other sources to back it up?
Are you accessing a translation of information in a language you can’t access? If so, try validating the information from multiple sources to ensure the translation isn’t intentionally or unintentionally misleading.
Describe the source ✏️
Are you being transparent in how you’re describing the quality and credibility of the source? After answering the questions above, what does the reader need to know about the source to enhance the credibility and transparency of the analysis?
Are you being sufficiently descriptive about the information the source provides? Don’t make intelligence consumers dig into hyperlinks or dive into footnotes to understand the main points you’re trying to get across.
Did you provide a source summary statement? This helps the intelligence consumer get a 360 degree view of the strengths and weakness of the collection of sources used in the analysis. It can discuss the strengths and weaknesses of the sources used, which are most important to informing key judgements, and what sources are corroborative or conflicting.
I am not suggesting these questions are going to save us from falling for every future pampered pope pic or AI-generated spam article.
But I am confident they can serve as a framework for thinking critically about the gobs of information we interact with on a daily basis. Consider them mini mental speed bumps to slow - or sometimes halt - how fast we fall for this kind of content. Because it appears more of it is coming. Buckle up.
Perspectives on Intelligence Collection Clark, Robert. Guide to the Study of Intelligence. The Intelligencer. Vol. 20, no. 2, Fall/Winter 2013.