Last week I wrote about our findings concerning people’s attitudes towards privacy and sharing of the types of personal data that the Beancounter (NoTube’s user profiling service) can collect. This blog post is going to focus on some early reactions to the Beancounter user interface itself.
The live version of the Beancounter wasn’t quite ready for testing last week, so Lora, Christoph, and I showed the Beancounter design mock-ups to our workshop participants (two women and five men, all aged in their twenties) to get some initial feedback. Based on these discussions we’ve drawn up the following list of our overall impressions, from which we can tentatively draw a few high-level conclusions. Given the small size of the group (and that all the participants are of a similar age), many of these comments are still somewhat speculative, and are intended to form the basis of possible future research directions, rather than as statements of definitive outcomes.
1. Personal analytics per se are of limited interest
The analytics part of the Beancounter was originally intended as a potentially interesting by-product of the user profiling, rather than its main focus. However, since data visualisation on the Web seems to be such a hot topic at the moment, we were keen to find out what participants thought about our ideas for presenting the data the Beancounter collects as a variety of charts and graphs. Whilst there were some aspects of the presentation they liked (or at least liked the intention behind), on the whole our participants didn’t really think they would find these kinds of analytics particularly interesting. Several mentioned that they thought the graphs and charts were a bit too ‘geeky’, ‘mathematical’, and ‘technical’. And, bearing in mind this feedback was from young academics working in the field of computer science, we might expect this group to be relatively favourable to this kind of presentation compared with the general population!
We had a short discussion about whether discovering surprising or interesting facts about the things you’ve done by seeing an aggregate of them (e.g. how frequently you watch the same things, or that you watch fewer educational programmes than you thought) could prompt you to change your future viewing behaviours. The group was roughly 50/50 split in their response to this: one person said they would probably be motivated to watch more educational programmes, another said it definitely wouldn’t change what she watched, even though she knew she was wasting time.
2. However analytics for social comparisons could be more interesting
As mentioned in part 1 of this blog post, the notion of wanting to compare your own TV watching habits with those of your friends, or even people you don’t know, cropped up several times during the workshop. This suggests to us that the ability to compare ourselves to others might be more of a motivational draw than simply seeing our own information in isolation, and so perhaps the Beancounter has some future potential in this direction.
3. “Recent activities” are useful as a memory aid
Participants consistently rated the “What you’ve watched or listened to in the last 7 days” list as the most useful part of the Analytics screen, because they saw it as a convenient ‘history’ tool for reminding them about a particular programme or music track that they’d forgotten.
4. People just want good recommendations and don’t care about the process
With this design of the Beancounter user interface, we wanted to test how much end-users would be willing or motivated to take part in the recommendations process by fine-tuning their own profiles (for example, manually adjusting the weightings of individual interests, as shown below). On the basis of the reactions of our sample group, it seems there’s currently little appetite for this level of interaction.
Our participants were of the opinion that there should be little reason for them to visit the Beancounter user interface, unless it was to delete some information. They expected to be able to correct a bad recommendation in the context of the recommendation itself (in which case they would assume that the Beancounter would automatically update). Therefore, whilst the Beancounter user interface might be a useful explanatory tool for making the workings of the Beancounter transparent, it seems unlikely to become a destination in itself.
5. Nevertheless a lightweight Beancounter user interface is still necessary
Given generally high levels of concern about privacy of personal data online, as discussed in part 1 of this blog post, we feel there’s still the need for some kind of lighter-weight version of the Beancounter user interface, to build trust and allow users to feel in control of their data.
6. So the challenge is to have the Beancounter whenever people need it, without overwhelming them
When developing this version of the user interface for the Beancounter we aimed to disclose more detailed information progressively, so that interested users could delve deeper should they wish to, whereas less interested users would not be overwhelmed with too much information. Our initial conclusion from this round of user research is that the next iteration of the user interface should be based on the assumption that the majority of users will only want the most basic scaled-back functionality (including the option to delete or hide specific data), so that nothing about the user experience should feel forced or obligatory.
A final thought
This last point ties in with the challenge identified at the end of the part 1 of this blog post: how to design a user experience that supports people’s need to feel in control of their data without making it overburdensome. Ultimately, the principle overall finding to emerge from our recent user research might be that such a lightness of touch should be applied across all aspects of the Beancounter front-end, and that focus on the interaction design would be more usefully applied to the presentation of the resulting recommendations instead.