10 Things Library Workers Should Know About the Internet's Invisible Architecture

Karen Lindell, ALA Public Programs Office

Perhaps one of the following questions has popped up from a patron at the library reference desk, or in your own mind while scrolling through social media: Why do ads for sunglasses appear right after I searched for flights to Hawaii? (Hint: You’ve got algorithms.) Why is the pope wearing a white puffer coat in that image I see everywhere? (Hint: It’s a deep fake photo.) Is artificial intelligence smarter than I am? (No.)

Navigating the internet and its invisible architecture — cookies, algorithms, large learning models (and other artificial intelligence building blocks) — can be bewildering. But a basic understanding of this underlying technology can help anyone respond to and interact with information from online sources in smarter, more beneficial ways.

To help library workers teach their adult patrons about these topics, as well as build their own knowledge about media consumption, the American Library Association’s Public Programs Office offers a free webinar series, “Media Literacy Education for Adult Audiences: Architecture of the Internet,” which focuses on topics in the recently published Media Literacy for Adults: Architecture of the Internet Programming Guide.

Topics of the five webinars are “Demystifying AI,” “Navigating Data Security and Privacy,” “Decoding Algorithms,” “Navigating the Attention Economy,” and “Exploring ChatGPT and Conversational AI.”

Here are 10 takeaways from the series:

1. Artificial intelligence (AI) encompasses a range of technologies.

Artificial intelligence (AI) is the ability of a computer or other machine to perform or simulate behavior normally associated with humans, such as writing, speaking, decision-making and problem-solving. The idea of AI has been around since ancient times — inventors have long attempted to create mechanical contraptions that mimic humans and their behaviors. But the term skyrocketed in popularity in 2022 with the release of ChatGPT, a chatbot from the company OpenAI that is trained to generate text in response to human prompts.

Some people tend to use “artificial intelligence” and “ChatGPT” interchangeably, but AI encompasses far more than one company’s chatbot program. AI technology is also used in everyday tools like Spotify and Netflix (to create personalized song and movie recommendations, respectively), Google (for foreign-language translations, to provide just one example), Waze (to analyze potentially traffic-prone roads), and Uber or Lyft ( to estimate arrival times and match drivers with passengers).

So even if you don’t use ChatGPT, or similar chatbots like Microsoft Copilot or Claude, you have very likely used AI in some form. 

2. AI has downsides.

As more people navigate, explore and experiment with AI, it’s important for them to think about the new technology critically, said Dr. Brandy McNeil, president-elect of the Public Library Association and deputy director of branch programs and services at The New York Public Library. Despite AI’s many benefits, such as taking on tedious tasks, and opening up new ways for people to communicate, the technology also has downsides.

For example, although AI might automate mind-numbing work that takes up an employee’s time, it will also likely result in job loss if humans are replaced with computers. Also, the experts who develop the technology can unintentionally embed their own or societal biases into their programming. For instance, an app trained to identify job candidates might leave out women or minorities if employees hired for the position have historically been white males.

AI is far from perfect, and without human interaction to check for errors, the technology can produce false or fake information, digital images or videos (the latter two are known as “deepfakes”). Copyright and consent are also concerns. AI chatbot programs like ChatGPT, for example, “learn” by digesting massive amounts of online text, images and data that might be copyrighted or used without a creator’s consent, such as photos and articles from a newspaper, or the text of an entire book. 

3. Protecting patrons’ data privacy is the proper thing to do.

Libraries collect a great deal of information about patrons — including names, addresses, emails, dates of birth and phone numbers — that generally falls outside state and federal regulations related to data privacy. Data about patron preferences and behavior, such as reading lists, might have even fewer legal protections. Despite fuzzy or nonexistent regulations, however, “protecting patron privacy is the right thing for us to do,” said Nathan Wittmaier, library systems manager for Mid-Continent Public Library in Independence, Missouri. Libraries should also consider patrons’ data privacy when using technology (such as AI) in their library services and programs.

In 2019, the American Library Association introduced amendments to the Library Bill of Rights that included a provision on privacy, with language that links intellectual freedom, privacy and security: “All people, regardless of origin, age, background or views, possess a right to privacy and confidentiality in their library use. Libraries should advocate for, educate about and protect privacy, safeguarding all library data, including personally identifiable information.”

4. Securing your library’s data systems is key to ensuring data privacy.

Libraries, just like any organization or individual, need strategies to prevent data breaches. Here are a few guidelines to implement with staff — and share with patrons — that might be familiar to you from protecting personal accounts and information:

  • Update passwords regularly (with long, complicated passwords that aren’t simply “password123”).
  • Use two-factor authentication for accessing technology, especially for administrative or other high-level access.
  • Avoid shared administrative accounts.
  • Limit the ability of staff to download sensitive information to physical hardware.
  • Consider how much information you need to know versus what’s nice to know, such as an adult patron’s three-month checkout history versus the person’s history from all the way back to their kindergarten years. 

5. Algorithms work by collecting, storing and analyzing user data.

Various definitions of “algorithm” might sound abstract, such as “a set of instructions with a given end result” or “a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer.” Algorithms are far from abstract, however, because they are responsible for what we see, hear and experience online, and guide many of the activities we engage in daily, from Google searching to social-media scrolling to buying stuff online. 

A search engine finds the best sites to direct you to during a search, for example, by using algorithms to analyze your past searches, as well as websites other people have linked to frequently. A commerce site suggests additional items you might want to buy based on algorithms related to what you’ve already purchased, or what people similar to you have bought.

Algorithms and artificial intelligence are integrated said Fiona O'Connor, senior services specialist in digital literacy initiatives at the Toronto Public Library, “AI uses algorithms to learn, and in order for AI to exist, it needs data.” 

Machines and programs powered by algorithms collect, store and analyze data from numerous sources, but mainly from people who might not be aware of the digital goldmine they provide: computer users such as you or anyone else who shares information online, whether it’s an opinion on social media, a phrase on a search engine, or a product in an online shopping cart.

Is human attention a commodity that can be bought and sold? This webinar explores the attention economy and how organizations use tactics in order to attract our attention to certain products and websites.

View all of the webinars in this series.

 

Conversational AI is a technology that enables machines to simulate human conversation. This webinar dives into what conversational AI is as well as the risks, benefits and considerations when utilizing this technology.

 

View all of the webinars in this series.

6. Algorithms are subject to bias.

Algorithms might logically seem “neutral” because they are based on mathematical steps and processes, but they are written and created for artificial intelligence applications, and as noted above, are thus subject to bias. Algorithms result from collecting and analyzing data that relies on patterns or information that can lead to harmful gender, age and racial biases. For instance, a facial recognition program used by a workplace, bank, police or border patrol station might not recognize the faces of people of color.

7. The goal of the attention economy is to generate profits.

The “attention economy” is a concept coined in the 1970s by Nobel Prize-winning social scientist Herbert A. Simon, in response to fears of information overload. He viewed attention as a scarce resource in an information-rich world, writing that “a wealth of information creates a poverty of attention.”  

The idea has become more popular, said Anna Kozlowska-Barrios, an associate professor and reference and liaison librarian at the University of Illinois Chicago, as more digital media tools compete for our attention by providing content through customized recommendations, subscriptions, ‘like’ buttons, following, friending and other methods intended to keep us attached to our phones.

The companies behind all these attention-grabbing tactics have one goal: to generate profits. By keeping us on our phones, they expose us to more ads and find out more about our individual preferences, which can lead to an eventual purchase or some other way to monetize our devotion to screens.   

8. Algorithms feed on our attention.

Algorithms are hungry … for our time and attention. Internet algorithms that power search engines or social media sites process massive volumes of data to deliver relevant information. For example, social media algorithms decide which posts or ads to display to us based on our preferences or our engagement patterns.

“It is … important to understand that these [algorithms] are designed and developed often by companies,” said Monya Tomlinson, a reference librarian at the Atlanta University Center Robert W. Woodruff Library. These companies’ lofty stated goals might be that they want to deliver the most relevant search result, but businesses also want to maximize engagement (i.e., attention) to keep people on their platform, which helps them make money.

9. Conversational AI tools like ChatGPT gobble up text to help them learn to sound like humans

“Conversational AI” refers to artificial intelligence technologies like ChatGPT and similar programs, as well as “virtual assistants” on websites, that are trained to simulate human conversation. 

Chatbots like ChatGPT, Microsoft Copilot and Google Gemini rely on technology called large language models (LLMs). An LLM is a large library of text material that uses books, articles, webpages and online conversations to discover patterns and rules of language.

Michael A. Spikes, director of the Teach for Chicago Journalism project and a lecturer at Northwestern University’s Medill School of Journalism, explained that the LLM drives a “transformer” that “breaks down textual patterns into tokens that generate strings of text by predicting meaningful relationships between words.” You don’t need to know too much about this process other than the fact that the “GPT” in “ChatGPT” stands for “generative pre-trained transformer.”

10. Consider reframing AI as a tool that helps you, not one that does the work for you.

Because artificial intelligence makes errors, and leads to problems with bias and copyright, Spikes suggest viewing artificial intelligence as “a tool that can help you get things done, not just do the thing for you. It can’t do everything a human does, but can help you do it faster.” For example, Spikes, says, when he writes an article, he might use ChatGPT to brainstorm ideas for a title, and he’ll tweak the title around the way he wants it. 

Kristen Calvert, events administrator for the Dallas Public Library, offered a few ways librarians can use AI “in the best way possible.” For example, she might ask ChatGPT to brainstorm ideas for the title of a workshop, or to generate a flyer that gives her ideas for visual layouts that might help inspire the way the actual flyer looks.

Calvert says it’s important to see if your organization has any rules or policies around using AI as an aid in your work. “It’s hard,” she said, because AI can be a helpful aid in your work when used right, but sometimes its use has negative perceptions. 

Related Links