How to Teach Research Skills in the Age of AI

Knowing how to do your own research is one of the most important takeaways from a good education.

It’s something I emphasized in my high school language arts classes, and something I value in my own life.

It’s also become kind of a joke.

That’s because these days, when someone says they “do their own research,” they’re often about to describe their journey down a rabbit hole of conspiracy theories, or their hours spent in an algorithm-based mental roll cage that reinforces everything they already believe.

But here’s the thing: doing your own research still is an important skill.

So how can you make sure that your students are doing research in the good, ongoing self-education way — and not in the conspiracy theory, rabbit-holey, terrible first date way?

This Problem Isn’t New

We can start by recognizing that a lot of the research-skill struggles teachers are dealing with are not actually new.

In fact, it actually makes sense to back up all the way to the time when Wikipedia started getting popular.

As a blanket statement based on my own observations, teachers were pretty negative on Wikipedia as a research tool early on in its life.

And I think that came down to two different issues.

The easy-button problem

One big teacher complaint about Wikipedia was that it felt like an easy button.

It replaced some of the actual work we expected students to do, and it replaced some of the mental exercise they were supposed to get from their assignments.

The gray-area problem

The other issue was that Wikipedia was an early example of the gray area of information quality.

It was nowhere near as gray as the gray areas we’re dealing with today — but it was an early example where information was getting crowdsourced, there wasn’t necessarily a professional editor in charge, and there wasn’t a publisher whose reputation would be at stake if the information turned out to be wrong.

Over time, it feels like teachers have become more accepting of Wikipedia as a research tool.

After all, Wikipedia can often give you a pretty good starting point when you’re researching something.

It can be wrong. It’s a little looser on fact checking than a traditionally published encyclopedia might have been, back in the day. But at the same time, it serves a purpose: it can keep pace with new topics much more quickly than a traditional encyclopedia ever could.

Overall, it’s become more accepted that Wikipedia can be a solid part of your research — as long as you understand what part of the research it works for.

AI Is Wikipedia on Steroids

AI can often give you a tremendously good answer, you can ask a follow-up question tailored to your exact needs.

At the same time, those two problems introduced by Wikipedia — the easy button and the gray area of information quality — have continued to today. And they’ve gotten more supercharged as technology has gotten more supercharged.

Google was even more of a gray area, and even more easy-button-ish, than Wikipedia.

And now, with AI? We just don’t know.

To be clear: an easy button is not always bad.

Google didn’t ruin research. And AI (probably) isn’t going to ruin research either.

But it definitely is going to change it.

One useful way to think about it: using AI is like “Googling backward.”

Google reaches out from what a person already knows and finds examples of it on the internet. It also ranks those examples by relevance and quality.

AI is better at going out and gathering information you don’t know and pulling it toward you.

It not only condenses the process of looking something up in the first place, it condenses the process of following up on whatever you learn.

You can ask follow-up questions, clarify, track down something you can’t quite name.

AI can meet students where they are… potentially better than Google. But the level of personalization with AI raises additional concerns.

With Google, you don’t have to worry as much about hallucinations being built into a conversation and reinforcing a wrong belief in an extremely authoritative-sounding and responsive voice, which is one of the current issues with AI.

We’ve also seen how certain aspects of technology can spin out of control in ways we don’t anticipate.

I was in my early twenties when Facebook came on the scene, and it just seemed like a great way to reconnect with high school friends.

I could have never anticipated what engagement algorithms would go on to do to our minds and to society.

So I have no confidence in my own ability to predict the technological future — and only limited confidence in other people’s ability to do so. We’re hearing a wide variety of predictions right now, and the majority of them will turn out to be wrong.

Am I saying this to manage my own anxiety about the future? Maybe.

But that’s why I try to focus on helping teachers in their classrooms right now.

Here are a few basic research principles to discuss with your students.

These are not specific to AI, but they are still relevant in the age of AI — maybe even more so.

1. Know Your Starting Point and Your Destination

The first thing anybody needs to do when researching a topic is to be aware of what level of expertise they currently have and what level they’re trying to get to.

There are times when an easy button is fine, because there’s an amount of friction that will make a person say, meh, I don’t need to know this that badly.

It’s also perfectly fine to decide you don’t want to go that deep into a subject.

There are certainly topics worth reading full books on. There are topics worth reading multiple books on. And there are tiny subtopics that people spend their whole careers researching.

All of the examples above are research.

But quickly looking something up on Google or asking AI is also research. You don’t need to become an expert on everything.

So your first research skill is: how much information do you already have, and how much more do you need?

2. Evaluate Your Sources

The next research skill is figuring out what sources you’re going to consult and how reliable they’re likely to be.

Ask yourself: is there a reason to think this source is either biased or misinformed?

It feels like we go through this cycle with each new technology.

Google linked us to an unheard-of amount of information, and ranked this information by quality and relevance. But over time, it turned out that ranking algorithms could be messed with.

And they’re often manipulated by people who are either selling things or pushing biased information, especially on controversial topics.

Now we have AI, which always sounds authoritative and pleasant and has perfect grammar. But we have no idea where it’s getting its information.

That doesn’t necessarily mean it’s completely unreliable, but there are new issues: AI’s source is the entire internet.

And we have no idea how it makes decisions about what to include in its summary of the entire internet.

We really don’t know yet whether AI will solve some of the issues of bias and misinformation on the internet, or whether it will simply inherit all of them.

We also know that AI sometimes makes mistakes in a bizarre, unprecedented way we’ve never seen before.

But also: AI is really easy to use.

3. If you want to tackle that gray area of information quality, you need to give up a little bit of the easy button. The way you do that is to check more than one source.

How many more? That really depends on how deep you’re trying to go.

There are people who have read every available book about one six-month period in history.

On the other end of the spectrum, simply taking the time to check one other reputable source is much better than going with the first answer you find.

If AI gives you an answer you’re not sure about, go to a different AI and ask the same question.

Or go to Google and type in your search term plus the name of a highly respected newspaper.

Or add National Geographic, or Encyclopedia Britannica — something that’s been around a long time and has a strong reputation for accuracy and thoroughness.

Is this a perfect strategy? No.

But comparing two or more reputable sources is always going to be more reliable than going with your first source.

The more boring and well-trusted the sources that agree on something, the more confident you can feel that you’ve got the basics right.

That’s especially true for controversial topics or topics where a lot of people are trying to sell you something.

When in doubt, treat AI the way you would treat Wikipedia.

It’s not a bad quick reference.

It’s not a bad shortcut.

It’s not a bad first source.

But as teachers have always emphasized to students that copying a Wikipedia entry does not mean you’ve done a research paper — neither does consulting only AI.