This reflection extends Behaviour & Responsibility — the ethics of systems that read what we would not otherwise say.
Artificial intelligence is becoming better at interpreting signals associated with emotion. Systems may analyse language, facial movement, tone of voice and behavioural patterns to infer whether someone appears frustrated, interested or distressed.
Used carefully, this may help organisations respond more sensitively. A service system may recognise that a customer needs human assistance. A learning platform may notice that a student is struggling.
But emotion is not a simple data point. People express feelings differently across cultures, personalities and circumstances. A quiet person is not necessarily disengaged. A smiling person is not necessarily comfortable. When systems treat emotional inference as certainty, misunderstanding can become automated.
There is also a difference between recognising emotion to offer support and recognising emotion to increase influence. A sales system may detect hesitation and change its approach. A platform may identify loneliness and promote content that increases dependence. An employer may interpret emotional signals as evidence of commitment or risk.
The power imbalance is significant. The organisation may know how the person feels while the person does not know how that interpretation is being used.
Emotional data deserves strong protection because it reaches beyond what people explicitly say. It attempts to infer their internal condition. People should know when emotional analysis is occurring and should not face serious consequences based solely on uncertain interpretations.
Human beings also need space to experience emotions without immediately turning them into data for intervention. Not every sadness requires optimisation. Not every hesitation indicates resistance. Sometimes emotions need to be heard rather than predicted.
Recognising an emotion creates a responsibility to respond with care — not an opportunity to exploit the moment when someone is most influenceable.