The discussion defines algorithmic harm as the use of algorithms to exploit consumer vulnerabilities, such as a lack of information or behavioral biases like unrealistic optimism. This goes beyond benign personalization to actively manipulate users into making poor purchasing decisions for products, services, or even financial investments.
Recommendation algorithms, whether for music, news, or social media content, tend to reinforce existing preferences, causing individual tastes to 'calcify.' On a societal level, this creates echo chambers where different groups inhabit separate factual realities, exacerbating political polarization and undermining social cohesion.
The guest argues that neither the U.S. nor Europe has effectively regulated the core harms of algorithms, with Europe's focus on privacy being insufficient. He proposes a new principle: a 'right to algorithmic transparency,' which would mandate that companies reveal the logic behind their recommendation and pricing systems.
A distinction is made between different types of algorithmic discrimination. Price discrimination based on wealth might be economically efficient and acceptable, while quality discrimination that sells inferior goods to less-informed consumers is clearly harmful. The key factor is whether the algorithm is targeting an informed consumer or exploiting a vulnerable one.
Keep pulling the thread on Cass Sunstein.