The Opt Out: The rewards and risks of lying to tech companies

Reclaim your online privacy: The Opt Out is here to help.

Algorithms are data-hungry beasts. These complex programs require high-quality data to function accurately. Insufficient or inaccurate data leads to flawed results.

My own experience with a malfunctioning algorithm highlighted this. My 2022 Spotify Wrapped inexplicably listed Peppa Pig as my top artist. The culprit? A week spent entertaining my niece with Peppa Pig songs. This seemingly minor incident revealed a broader issue: my music recommendations were flooded with children's music.

This led me to a realization: deliberately providing inaccurate data could disrupt the detailed profiles tech companies build. If platforms misinterpret our identities, they'll show irrelevant ads, and any data shared with third parties will be inaccurate. This "mistaken identity" acts as camouflage, protecting our privacy.

A Camouflage of Bad Data

Intentionally feeding algorithms inaccurate data is known as data poisoning or obfuscation. While sophisticated attacks require technical expertise and resources, individuals can employ similar principles for self-protection. Every online action—images viewed, posts liked, videos watched, music streamed, locations checked in—generates data used to create user profiles. Tech companies aim to understand us completely to predict our desires, ultimately driving targeted advertising.

Simple data poisoning involves using false information during account registration (name, gender, location, birthdate). Further obfuscation involves liking irrelevant posts, clicking unrelated ads, or playing unwanted content (videos, music, etc.)—even letting a platform autoplay content overnight. When prompted for explanations (e.g., returning an online purchase), choose "other" and provide a nonsensical reason.

Limitations of Data Poisoning

This approach isn't foolproof. Platforms often cross-reference multiple data points. Inconsistencies (e.g., claiming a California residence while consuming Wisconsin news) will be detected. Similarly, providing a wildly inaccurate birthdate that contradicts other data will likely be flagged. There's also the risk of needing to verify your identity if your account is compromised.

Continuously playing irrelevant content requires resources (electricity, high-speed internet). It also negatively impacts user experience. Relying on platforms for recommendations means inaccurate data will lead to irrelevant suggestions. In dating apps, for example, this could lead to unintentionally rejecting compatible matches.

Consistency is key. Sporadic actions won't significantly affect algorithms. Repeated actions are necessary to reinforce the false profile. Just as online ads cycle, so too must our data poisoning efforts.

The biggest uncertainty is the effectiveness of data poisoning. Studies suggest a small percentage (1-3%) of poisoned data can significantly impact algorithms. However, these are estimates. Tech companies constantly update algorithms, making them a moving target. The effectiveness remains largely unknown, as companies are unlikely to reveal vulnerabilities.

Does it Matter?

The uncertainty of data poisoning's effectiveness might seem discouraging. However, anecdotal evidence (Spotify Wrapped inaccuracies, YouTube's unusual recommendations, etc.) shows platforms aren't immune to inaccurate data. Research, such as the AdNauseam experiment, confirms the effectiveness of disrupting Google's profiling algorithm.

We are not obligated to comply with every online request. Data poisoning isn't dishonest; it's a user's attempt to reclaim control over their information. As Jon Callas of the Electronic Frontier Foundation stated, we have no moral obligation to answer questions tech companies have no right to ask.

The exact effectiveness of data poisoning is secondary. We know it has an impact. In a climate of weak regulation and corporate prioritization of profit over user privacy, users must employ all available strategies for self-protection.

Explore more PopSci articles.

The above is the detailed content of The Opt Out: The rewards and risks of lying to tech companies. For more information, please follow other related articles on the PHP Chinese website!