[Wishlist] Provide mechanism for guiding Do's learning

[David Brewer]

I just filed a bug (#328658) regarding Do's learning system. This is a related feature request.

I see a number of bugs from people that are confused about how to get Do to learn about their preferences when searching or choosing actions. It occurred to me that one way around this would be to provide a mechanism for nudging Do in the right direction.

Let's say you've just searched for a file, selected a specific action such as 'open with', and then selected the application you want to use. It would be great if you somehow had the option to tell Do "this is what I want you to do with this type of file from now on". This might be implemented by using the regular learning system, but just bumping up the score for the current course of action by a lot more than you normally would for one action. Or, similar to what was suggested in #257026, you might maintain some internal list of file names or types which are 'pinned' to certain actions.

I could see this being a special key combination you hit one you have selected the desired course of action, or perhaps a small icon that appears on the last pane. In the 'pinning' approach, this could look like a pushpin, and you could use it to pin or unpin the action depending on your needs.

[David Siegel (djsiegel-deactivatedaccount)]

The mechanism for guiding Do's learning is simply using Do. If you think that it is not learning well enough, or fast enough, the solution is to improve Do's learning features, not to add additional mechanisms for the user to explicitly affect the learning system.

On the other hand, if we cannot sufficiently improve the implicit hints and Do's ability to learn user preferences without being told, *and* we are able to determine that the added complexity you suggest would result in positive net user benefit, then we could consider explicit learning hints.

[Jack Senechal (jacksenechal)]

I used this feature frequently in Quicksilver. It came in quite handy at times. I've recently wished for it often, since Do keeps unlearning certain actions (which is, of course, a separate bug).