Too, too much here to possibly summarise neatly, but I can only suggest that you read this apparently daunting 66 page report which is in fact incredibly easy to read and follow and makes lots (and lots) of valid and relevant points. There’s just sooo much that big data can tell us and be useful for, much of which we’ve covered already in the #LAK12 mooc, but with equal amounts of health warnings and concerns. So, here’s a random selection of snippets and quotes I picked up:
- are we in danger of manipulating customers (students) to behave in particular ways that suit us and not them?
- can we always trust what the data tells us? We need a way of modelling that uses context as well as just numbers (‘Whenever you do statistics …you find spurious correlations—apparent relationships or proximities that do not actually exist.’)
- It’s all too tempting to impose a view that we think is there (or should be there) (‘Cleaning the data’— deciding which attributes and variables matter and which can be ignored—is a dicey proposition, because ‘it removes the objectivity from the data itself. It’s a very opinionated process of deciding what variables matter.’ and ‘The prejudices of a society are reflected in the algorithms that are searched’)
- Correlation does not imply causality much as we may want it to (‘Causality requires models and theories—and even they have distinct limits in predicting the future. So it is one thing to establish significant correlations, and still another to make the leap from correlations to causal attributes’)
- Being correct doesn’t always matter – you only have to predict what people think is going to happen to influence behaviour…
- privacy concerns (‘Is personalization something that is done to you or for you? ….. consumers have far less knowledge of what is going on, and have far less ability to respond’)
- is it in our interests as businesses(/universities) to allow customers(/students) to know too much ? Failure of a service can lead to repeat business, early success can move the customer(/student) on too quickly (‘intelligent uses of Big Data are frequently resisted. Why? Because health insurers, pharmaceutical companies and patients often believe that their self-interests will be harmed by the collection and use of data’)
- efforts to protect privacy by depersonalising data may be naive (because we’re dealing with Big Data, after all)
- A really interesting car insurance example made me consider the following: will there come a time when students, armed with their own data sets, can ask universities to bid for them as students because they are low risk (and low cost)? Where could that leave the high risk (high cost) student….?
- How far can and should we go in presuming future behaviour if it means that we impose limitations/constraints on how we support students? (‘probabilistic cause … should require greater transparency’ or is that just too much information for most people?)
One of the best reads so far, and highly recommended.