A Smart Home That Stopped Feeling Smart
Daniel, 68, lives alone and has gradually adopted connected devices. His smart thermostat learns his routine and now pre-warms the house at 6:10 a.m.; after his daughter and grandchildren stayed for a week, it began heating the house at 5:30 a.m. and to 23 °C, and it has not fully reverted. His bank app blocked a £240 payment to a shop he has used for years, then allowed a £900 transfer he did not recognise until he called. His phone's face unlock stopped working after he started wearing reading glasses. He likes the convenience but worries he is losing control. He asks his nephew, who knows a little about technology, what is actually 'intelligent' here and what he should do.
1. Classify the thermostat, the bank fraud check and face unlock as rule-based, learned, or hybrid, and explain the everyday evidence that supports each classification.
Read the worked answer
Thermostat - learned. It builds a schedule from occupancy and temperature data; the change after the family's visit is the tell. No one reprogrammed it, so the 5:30 a.m. and 23 °C behaviour came from recent patterns, and it persists because the model has updated.
Bank fraud check - hybrid. Rules (country blocks, velocity limits) sit alongside a model trained on past transactions that scores each payment. The £240 block is a false positive from an unusual pattern; the £900 passing shows the score is probabilistic, not a guarantee.
Face unlock - learned. A neural network maps facial features to an embedding; it is not matching a stored photo against rules.
Everyday evidence is adaptation without explicit reprogramming. The opposite - a fixed schedule Daniel set by hand - would be automation.
2. For each problem, suggest one practical action Daniel can take, and explain the trade-off between personalisation and control.
Read the worked answer
Actions:
- Thermostat: switch off learning/adaptive recovery and set a fixed schedule, or reset the learned schedule and re-enable learning only when the house is empty. A geofence can help.
- Bank: enable instant notifications, set a lower card limit, review payees, and call the fraud line about the £900. Do not treat a passed payment as evidence of safety.
- Face unlock: re-enrol while wearing glasses in different lighting, keep a PIN/biometric backup, and consider whether face unlock is worth the failure mode.
Trade-off: personalisation needs data about Daniel's habits, which improves convenience but reduces predictability and shares information with vendors. The right balance is the one where he can explain and reverse a decision. Defaults should be inspectable, and he should know how to reset them.
3. The bank allowed a £900 transfer Daniel did not recognise. Who is responsible, and what should he do now?
Read the worked answer
Separate three questions: was the payment authorised, was it correctly authenticated, and was it detected as anomalous?
- Daniel should call the bank immediately, report the £900 as unrecognised, ask for a freeze and a transaction trace, and change credentials if phishing is possible.
- The bank's liability depends on whether strong authentication was used and on local payment rules; the fact that its system allowed the transfer does not make Daniel responsible by default, but delayed reporting weakens his position.
- The £240 block is a false positive, not proof the system works. Both events show the detector is calibrated for some patterns and not others.
He should ask for the reason for the block and the missed fraud in plain language. A human review and a complaint route matter more than the app's 'smart' label.