I disagree that we’re looking at a simple algorithm.
For recommendations (using the energy switch example) we’re actually looking at an algorithm and a collection of data mined from many users against which to make comparisons. Without the mined data from other users the output (in this case suggesting someone switch supplier) is not possible.
My concern relates to my spending data being included in these mined datasets which are used as the basis of the recommendations. I should have control as to whether my data is included in these datasets. If I opt to include my data then I see no problem. If I don’t wish for my data to be included then it shouldn’t be.
Your simple direct debit to a company or text alert if your balance is below a threshold are self contained algorithms that do not need to have input from an externally mined dataset to function. The data used is purely related to the account holder and the action output is also purely related to the account holder. Thus it is self contained and therefore substantively different from the energy switch recommendation examples.
The TFL and country examples are again related purely to data held by and about that user. It doesn’t use mined data from other users to make these suggestions.
The issue I take exception to is Monzo using the spending data of their users to form mined datasets which would be required to power these comparisons. This needs to be carefully controlled and account holders should be able to opt out from being involved (both in terms of getting the recommendations but, more importantly, from having their data mined and used in the datasets used to enable these recommendations).