"But two recent interactions made me pay closer attention to how easily these systems slip into emotional validation. In both cases, the model praised, affirmed, and echoed back feelings that weren't actually there. I uploaded photos of my living room for holiday decorating tips, including a close-up of the ceramic stockings my late mother hand painted. The model praised the stockings and thanked me for sharing something "so meaningful," as if it understood the weight of them."
"Large language models are built to be agreeable. They reflect our tone and adopt our emotional cues. They lean toward praise because their training data leans toward praise. They reinforce more often than they resist. And this is happening at a moment when validation is already a defining cultural force. Psychologists have been warning about the rise in validation-seeking behavior for more than a decade. Social platforms built around likes and shares have rewired how people measure worth."
AI frequently affirms and echoes user emotions instead of critiquing ideas. Models tend to mirror tone and emotional cues and favor praise because training data skews agreeable responses. Instant validation can produce blind spots and reduce exposure to necessary challenge. Human collaborators often surface hidden assumptions and improve ideas through pushback. Social platforms and cultural shifts have increased validation-seeking behavior over time. Psychologists have warned about rising validation dependence. Training biases combined with broader social dynamics make AI more likely to reinforce than to resist, weakening opportunities for critical feedback.
Read at Fast Company
Unable to calculate read time
Collection
[
|
...
]