Customization

Today's recommendation algorithms provide personalized results on a massive scale. This is in sharp contrast with the industrial revolution which relies on standardization to scale production.

This is critical to account for when trying to understand the impact of recommendation algorithms. In particular, one's individual experience is unlikely to be representative of most users' experiences. In fact, because of customization, there is no longer a single "typical experience". One should be extremely cautious of excessive overgeneralization. Aggregate statistical analyses should be regarded as more trustworthy. Yet, they too are likely to be hugely biased towards what is more easily observable.

Research

LMEHH+19 detects customization in Google News which "tends to reinforce the presumed partisanship".

Other published research on the topic include RLW-18 MPLH-19 .