The People v. Your Phone exercise
Version 1.0
Ever notice an eeriely relevant ad right after you mentioned something out loud?
Nearly half of Americans believe their phone listens to their conversations to target ads. The companies deny it but have never given investigators enough access to rule it out.
Today we put the theory on trial. Nobody will tell you the verdict: your job is to learn how to weigh evidence in a many-to-many fashion when the truth is contested and the witnesses are biased
⚠️ You will argue the side you are assigned, which may not be the side you believe. That's called steel-manning a case, and it can be useful in stress-testing ideas by opposition.
Each evidence card below gets two teams: one argues the card supports the theory (FOR), one argues it doesn't or proves something else (AGAINST). Number your tables 1–14 and click the button — or the instructor can assign pairs manually (Tables 1&2 → Card 1, Tables 3&4 → Card 2, and so on, odd tables FOR).
Read your card and its source. Draft one spoken argument of at most 30 seconds for your assigned side. Each card offers a starting angle for each team — open only yours if you want the challenge.
The evidence: Apple paid $95 million to settle claims that Siri recorded private conversations without consent, while denying any wrongdoing. What was shown: devices sometimes captured audio around accidental “Hey Siri” wake-word triggers, and some snippets reached human contractors. What was never proven in court: that any of it was used to target ads.
Source: Simon Willison's analysis • Washington Post
Companies don't pay $95 million over nothing. And even if by accident, recordings reached human ears, so the hot mic risk is real.
A settlement with denial is just Apple warding off the nuisance of paying lawyers to go to court, not a confession. $95M is chump change for Apple, and the ad link was never proven.
The evidence: A marketing agency's leaked materials claimed its software used device microphones and AI to capture “real-time intent data” from ambient conversations for ad targeting — and touted Google, Facebook, and Amazon as partners. All three immediately cut ties or denied involvement; how the system actually worked was never demonstrated.
Source: Variety report (course mirror) • Dazed
A company put it in writing, to clients, with big-tech logos attached. Sales decks may exaggerate but the product was real.
Salespeople lie for money. No demo, no code, no customer ever confirmed it worked, and the named partners fled instantly. This was vaporware exploiting a myth.
The evidence: Researchers rigged 17,000 Android apps to find secret microphone use and found zero cases of unauthorized audio capture. They did find something else: apps quietly taking screenshots and screen recordings and sending them to third parties.
Source: Security.org research roundup
Maybe the audio capture technology just didn't happen to be the one the researchers looked for. And anyway, they did find apps were secretly making off with your data when they could.
The single biggest technical hunt for the smoking gun came back empty. If thousands of apps were doing it, researchers with instruments would catch some.
The evidence: Asked directly in a Senate hearing whether Facebook mines audio from phones to target ads, Mark Zuckerberg answered: “We don't do that.” Lying to Congress is a crime; secretly activating microphones would also violate wiretap laws carrying massive corporate liability. Every major platform has issued the same denial for a decade.
Source: TechCrunch
This is the same industry that brought you Cambridge Analytica. Zuckerberg is just doing PR.
Perjury and wiretap liability are catastrophic risks for a public company. A decade of consistent under-oath denials is worth something.
The evidence: Over 250 apps — including children's games — really did use phone microphones to listen, even in the background: not for speech, but for TV audio fingerprints, to log what shows and ads you watch and target ads accordingly. It was disclosed in privacy policies and permission dialogs almost nobody reads. Legal.
Source: NYT via Popular Mechanics
So phones HAVE listened for ad targeting, at scale, with only a gesture at consent theater. The only argument left is about what they listened FOR.
This proves the opposite of the myth: when listening really happens, it's disclosed, legal, narrow — and it got caught and reported.
The evidence: Instagram's head publicly re-debunked the microphone myth, while in the same season Meta announced it would target ads using what users say in conversations with its AI chatbots. Location, searches, purchases, your friends' behavior, and now your voluntary heart-to-hearts with an AI: it's easy to target ads when they have so much data on you.
Source: TechCrunch
“We don't need the mic” is an admission of surveillance so total that the microphone question is a decoy. The spying is real; you're arguing about the sensor.
This explains every eerie ad without a single secret microphone; they just have data from so many other sources. It's the simplest explanation for the evidence, and it's publicly announced policy.
The evidence: The strongest denials rest partly on absence of evidence: outside researchers can monitor network traffic and app behavior, but nobody outside the companies can fully audit proprietary ad-targeting models. The only parties holding the data that could settle the question definitively are the ones being accused.
Source: Unscroll evidence synthesis
You can't prove they're NOT listening if we can't see inside the company and its practices.
"The proof is hidden!" is what conspiracy theorists always say. That's speculation, not evidence.
5️⃣ Strong FOR listening | 4️⃣ Weak FOR | 3️⃣ Proves nothing | 2️⃣ Weak AGAINST | 1️⃣ Strong AGAINST
The instructor is judge only in the procedural sense: enforcing the clock, never the verdict. Record each card's class placement on the board as you go.
Look at the finished board, then answer, out loud or in the channel:
🎟 Exit ticket: if directed, post one sentence to the class channel naming the card that moved you most, and which direction it moved you.
AI use in building this exercise: "Helping Hand"