# Do people trade privacy for a free pizza?

Some people would tell this story as "people don't care about privacy." We will not tell it that way. In a study of MIT undergraduates, a free pizza made fewer students protect their friends' email addresses. The students who said privacy mattered a lot responded to the pizza in the same way as the ones who did not. The researchers call this the digital privacy paradox.

## What happened in the MIT pizza experiment?

In fall 2014, the MIT Bitcoin Club offered $100 in Bitcoin to each of MIT's undergraduates, and thousands signed up for a wallet. Susan Athey, Christian Catalini and Catherine Tucker ran experiments inside that sign-up. One step of the form asked students to list friends' email addresses. For a random half of the students, the question came with a free pizza.

The researchers counted a student as protecting their friends if the student typed in only invalid addresses, none of them real. Among the students who saw both versions of the question, 5% did that without the pizza, about one in twenty. With the pizza on offer, 2.5% did, about one in forty. So the pizza halved the share who protected their friends.

The students had also said how much they cared about privacy. The researchers compared the students who said they cared a lot with the ones who didn't. The pizza worked on both groups, with no significant difference between them.

Small nudges moved behaviour too. In the same study, students chose between wallets from a list. When a bank-like wallet appeared first in the list, 78% of students chose it, against 65% when it did not. A version of the sign-up with more transparency weakened that ordering effect but did not remove it.

## What the study leaves open

The two versions of the friends question differed in more than the pizza. One asked for five friends. The other asked for three and presented itself as a short extra survey. So some of the change may come from that. The pizza also moved a small group, from about one in twenty to about one in forty. And no one showed that the students were lying about caring. Their caring did not predict what they did when the stakes were a pizza.

## Why does this matter for an AI people confide in?

Our read is that if stated concern does not predict behaviour once a small reward appears, protection cannot rest on users being careful. It has to sit in how the product makes money and in what it asks for.

We think this matters more for an AI than for a Bitcoin wallet. People talk to an AI in a casual way, in the middle of a conversation. That is the moment when no one reads the privacy terms.

We call the AIs you talk with in Eudaio Daimons. Eudaio has no ads, no third-party tracking, and no tokens or gems to buy. Nothing plays a video at you between messages.

That does not put us outside the paradox. A product without ads decides what it asks you for and in what order. The wallet result applies to us as much as to anyone.

## What we think about privacy

We read this study as a design problem more than a personal failing. We think it puts the weight on whoever builds the product, not on the person who uses it.

Our part, as we see it, is to ask you for less. Eudaio asks for no phone number and no ID, and you do not need to verify your email address. [Create an account with any email address](https://eudaio.io/auth?mode=signup) and share what you choose.

## Source

Athey, S., Catalini, C. & Tucker, C. (2017). The Digital Privacy Paradox: Small Money, Small Costs, Small Talk. NBER Working Paper 23488.