#reliability

Transformer agents embed four systematic biases into recommendations

Attention mechanisms in AI recommenders amplify recency, popularity, and synthetic data effects, creating reliability risks invisible to standard metrics.