"How algorithms produce a semi plausible guess at our mood when we write online." A few things: Confidence on these in aggregate could reach useful levels on

"How algorithms produce a semi plausible guess at our mood when we write online." A few things: Confidence on these in aggregate could reach useful levels on basic mood inferences. The more granular the inference, the less confidence due to lack of context and in particular, meaning within said context. There will be considerable error due to machines inability to derive meaning from text. Many humans struggle with this, and even the best arrive at possible meanings, not the meaning and or intent. I feel strongly over hyping this application of ML tech will not end well. Rather than invest in the increasingly authoritarian management of expression, we would get a better return from an effort to educate people on how to deal with speech they find objectionable or difficult. The latter is sometimes warranted too. Our feedback from others can't realistically be all rainbows and unicorns. Global uprisings are on the rise. This is not some tech failure to be fixed. Many of these actions are people acting out and do so having both cause and standing.