← Back

Visible & Invisible AI

A learner can get through a whole day, pass through more AI than they could name, and not notice any of it. Because AI mostly arrives inside something else. It’s tucked inside software a school already uses. A phone in someone’s pocket. A music platform deciding what plays next. It reads a job application before a recruiter does. It puts a flag on a student, and that flag follows them into next year. It shapes what a family pays. Usually there’s no little sign that says “AI did this.”

The goal of this Compass Collaborative resource is to make the invisible AI visible, so you can make sense of what’s already around you before you decide what to do about it. None of the nine below is good or bad on its own. A screen reader and a screening tool can both be the right call or the wrong one. What changes the outcome is whether someone chose it on purpose, and whether the person on the other end knew. Some forms of AI are easier to spot than others. We’ve ordered these roughly from most visible to least. At the top, you choose to use the AI, and by the bottom, you mostly just encounter its decisions.

Visible AI to Invisible AI

Visible AI → Invisible AI

  1. Talking

    Has a conversation and answers questions.

    You open it, you ask, it answers. For example: a chatbot in one tab, a tutor working through a proof in another, a companion that remembers what you said yesterday. It can be remarkably patient. It can also sound remarkably confident while being completely wrong.

    Try this: Ask it where it got its information, then go check whether the source is real.

  2. Making

    Creates images, video, music, and writing on request.

    You describe what you want and, suddenly, there it is. For example: a PowerPoint illustrated in under two minutes, a song built from a hummed melody, a first draft where there used to be a blinking cursor. It’s great for getting you past the blank page. But it also hands everyone else something that looks a lot like yours.

    Try this: Write a first draft. Ask AI for a critique. Then edit again.

  3. Helping

    Works alongside a task to make it easier.

    This kind of AI can disappear into the task itself, so it’s easy to stop noticing. For example: captions under a lecture, a menu translated at the table, a screen reader keeping pace with the page. Sometimes AI makes a task easier. Sometimes it’s what makes the task possible at all.

    Try this: Go through the accessibility settings on a device you already own before you buy anything new.

  4. Embodied

    Shows up in something you can talk to, wear, ride in, or hold.

    For example: a toy that answers to its name, smart glasses that give directions, a car that parks itself outside school. When AI has a voice or a body, it feels less like software and more like a friend. That makes it intuitive, and easy to get attached to. And hard to tell when it’s listening.

    Try this: Find the microphone setting before you hand it to a learner.

  5. Agents

    Takes a goal and handles the steps for you.

    For example: a trip booked overnight, groceries reordered without a list, a calendar rearranged while you slept. You see what it hands back, but not the steps it took to get there. It runs the whole errand and shows you the receipt. What’s harder to understand are all the little decisions it made on the way there.

    Try this: Set the limits before you hand off the task.

  6. Choosing

    Decides what shows up next.

    Nothing announces itself here. Things just arrive in an order chosen for you. For example: the next YouTube video already playing, an AI summary sitting above the search results, an ad that weirdly matches last week’s searches. You mostly get what you wanted, but you never see what got left out.

    Try this: Turn off autoplay, and take one pass through the recommendation settings on whatever your family uses most.

  7. Judging

    Scores, screens, or evaluates people or their work.

    For example: an essay scored before the teacher sits down with a red pen, an application screened before anyone on the hiring team reads it. Fast can be useful. But scoring wants neat categories, and one rubric for everybody leaves little room for the things that make a person interesting.

    Try this: Ask whether a person reviewed the result, and what happens if you disagree with it.

  8. Watching

    Keeps track of activity and surroundings.

    This kind of AI runs in the background whether or not anyone is at the other end. For example: a school laptop that flags certain words, a family app that shows the walk home, a doorbell that recognizes a face. It can give peace of mind on one side of the screen. On the other side, it’s a record.

    Try this: Ask your school what its monitoring software flags, who sees the list, and how long it’s kept.

  9. Sorting

    Groups people so something else can happen next.

    You may never see how people are sorted. You just meet the consequence. For example: a student lands in a specific reading group, a service becomes available once someone crosses a threshold, a flag appears in an office system that another person will use later. It can feel like being understood. It also means next year’s teacher meets the label before they meet the learner.

    Try this: Ask how someone gets out of a group once they’ve been put in one.

Invisible AI

Invisible doesn’t mean sinister. Most of what’s on this list is useful, and some of it is the reason a person can take part at all. But you can’t shape what you can’t see. Noticing the AI around us allows us to better shape it, question it, or opt out of it. There’s no one right relationship with this stuff. Start with the learner in front of you.

AI Guidance by Age