Algorithmic filtering amplifies preexisting biases rather than causing them

Evidence

In short, these findings illustrate that, although social media may not be the root cause of our problems, its algorithmic filtering can create an echo chamber that amplifies and exacerbates existing biases, especially amongst those individuals who may be more open to considering alternative or extremist content in the first place. — Sander van der Linden, Foolproof, 414

Categorization

• XREF: Directly extends the filter-bubble and prior-susceptibility themes running through earlier chapters of this book. — connects to: Bias-correction algorithms can equalize treatment while masking latent power to discriminate; Predictive algorithms cannot cleanly label predictors legitimate or biased proxies — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1184; Partial debiasing can make bias problem worse by hiding it — Brian Christian, The Alignment Problem, 210 (+5 more)

Broader Connections

• Bias-correction algorithms can equalize treatment while masking latent power to discriminate #complicates (0.85)
• Even well-intentioned technology can embed hidden functional biases #corroborates (0.85) — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1143
• Debiasing algorithms risk exacerbating bias without a careful implementation framework #corroborates (0.82) — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1223
• Partial debiasing can make bias problem worse by hiding it #contradicts (0.80) — Brian Christian, The Alignment Problem, 210
• Predictive algorithms cannot cleanly label predictors legitimate or biased proxies #expands (0.78) — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1184
• Machine-learning bias often stems from unrepresentative training data #qualifies (0.75)