An independent study of what people actually want from AI
An independent study of what people actually want from AI
Methodology
We believe methodological transparency is part of what makes independent research trustworthy.
This page outlines our study design, analysis framework, budget expenses, and what we'll deliver publicly.
Our Study Design
Data Sources
Main Project
-
20 minute online survey
Stretch Goals
-
Qualitative interviews
-
Social listening
Sampling Criteria
-
N=4000 Nationally representative US sample
-
Must be at least 18 years old
Limitations
-
Self-reported data
-
Cross-sectional design
-
US-only sample in the primary study, international perspectives are a stretch goal
Research Questions
1. How Are People Engaging With AI Today, And Why?
Some people avoid AI. Some are curious but cautious. Others experiment, use it selectively, or rely on it in parts of work and daily life. We will map what people are doing, where AI does and does not fit, and why, without assuming that more use is always better.
2. What Do People Actually Want AI To Help Them Accomplish?
Not simply what AI can do, but what people want it to do for them. We will identify the outcomes people value most, including more time, less mental load, expanded opportunity, and help with problems that otherwise feel out of reach.
3. When Does AI Create Friction, And When Does It Create Empowerment?
Empowerment means AI genuinely helps people and makes them more capable. Friction appears when AI creates pressure, unhealthy dependence, or eroded skills. We will measure both sides, including when benefits and drawbacks coexist.
4. What Gives People Pause, And Where Do They Want Clear Boundaries?
We will examine concerns about accuracy, privacy, bias, job effects, human connection, over-reliance, and AI acting beyond its proper role.
5. What Makes AI Use Feel Worthwhile?
Value is not just speed. We will identify which uses feel valuable enough to justify the effort, cost, data, risk, and oversight involved. We will also identify what people need, including reliability, privacy protections, and transparency before the tradeoff feels worthwhile.
6. What AI Relationship Profiles Emerge, And What Guidance Fits Each One?
We will create profiles by combining two dimensions: how far AI has reached into someone's life, and whether that relationship is actually helping them. These profiles will inform where AI may help and where caution is warranted.
Support The Project
How We Will Make Sense Of The Findings
Two things get measured separately: how far AI has reached into someone's life, and whether it's helping or hurting them. Almost everything below is about how those two fit together.
1. Measuring the two dimensions
Before anything else, we will use reliable scales to measure how far AI reaches into a person's life, and whether it's helping, keeping them in control, and leaving them more capable. We test these to make sure they hold up: that they measure what we say they measure, and that people answering honestly get consistent results. Answers Questions 1 and 3.
2. Grouping people into profiles
Rather than sorting people into categories we invented in advance, we let the data show us which patterns actually exist (a technique called cluster analysis). People whose answers resemble each other end up in the same group. These groups become the AI relationship profiles. Answers Question 6.
3. Looking at how the two dimensions interact
This is the center of the study. A deep, personal relationship with AI is empowering for some people and costly for others. A purely functional relationship can go either way too. We examine which combinations show up, how common each one is, and what separates the people who benefit from the people who don't. Answers Questions 3 and 6.
4. Finding out what predicts what
Simple averages hide a lot. Using statistical models (multivariate regression), we can ask questions like: does feeling understood by AI actually predict relying on it more, once you account for age, job, and how much someone uses it? This is how we separate real drivers from coincidence. Answers Questions 2, 3, and 5.
5. Weighing the trade-offs
Most real AI use is mixed: it saves time and costs a skill, or offers support and takes some independence. We measure the good and the bad separately so people can see both, and we look at which trade-offs people say are worth it and which cross a line. Answers Questions 4 and 5.
6. Comparing across people and settings
We compare profiles by demographics, profession, and industry (education, healthcare, consulting, and others) so the findings say something useful about your situation, not just the average American's. Answers Questions 1 and 6.
Budget & Research Expenses
See our policy page for further detail on funding, scope, risks, and refunds
Publicly Available Deliverables
-
Research report on what people actually want from AI
-
Video and Infographic series to make research findings accessible
-
AI Relationship Profile tool to help you understand your own AI expectations and what would make it work for you
-
Data dashboard for further analysis
Support The Project