A genuine use of machine learning

AI winter apart, as I mention in my previous post “Do you want to help us enhance the product experience?“, I suggest one more use of machine learning – learning the preference of users about various GUI controls.

Deriving genuine use cases for a GUI is an exciting possibility, isn’t it? Has anyone tried this before?

Do you want to help us enhance the product experience?

A lot of software nowadays ask this question. They collect anonymous crash reports and send it to the developer.

There are two other ways I can think user experience can be enhanced.

One I discussed in previous post titled “One Way Help“. If automatic crash logger sends Severity 1 issues, such an “Inverse Help” may send in Severity 2, 3 and 4 bugs.

The other idea is about the GUI layout. Outlook is the worst example of organizing menus. Just try to “Search” for the “Search”!

I don’t know what experts of GUI design say. I can see menu organization as a learning exercise. When I say “learning”, I mean “machine learning”.

A Hidden Markov Model or an Artificial Neural Network that learns the sequence of GUI controls pressed can get a gist of how the software is “controlled”.

Such learning can periodically and with explicit permission of the user, be sent to the software developer for realignment of frequently used GUI controls.

Based on various conventions (like “the top 10 controls should become buttons”) next version of the software can become very user friendly.

What do you say?

One Way Help

If readers of this blog remember, we had discussed about one ignored part of normal menus in applications – “Open/Save/Save as/Move to“.

Today, I was talking to one of my friends in marketing and I remembered my earlier wonder about why another part of UI is so much ignored.

That is “Help” menu!

Frankly, I think this is the LAST menu in the list of love of software developers. I am not sure whether the history of “Help” menu would go much deeper in time.

Typically such help is directed from the developer to the user. I will call it “One Way Help” – from developers to the users.

Ironically, users somehow live with the product. It is developers who need more help from users! Here is how.

If you are marketing a software, your biggest problem is “What do my customers want?”. [Even if you are doing FOSS, you still need to market your software.]

I have seen only Mozilla Firefox coming nearer to seeking a feedback through Help menu. That too is so buried and difficult to use, I wonder whether “its potential is realized” (that is “anyone uses it at all”).

****

These days word “Help” is substituted by a “?”.

In my humble opinion, bigger “!”  menu should be displayed beside it. Clicking on it should lead to a very user friendly plug-in to register feedback – complaints and more valuable, suggestions.

That will answer “What do my customers want?” straight from the horse’s mouse. [No typo.]

How could it be very user friendly?

Say you clicked on the “!”. This should

1. Open a screen snapshot similar to “Print Preview” but removing all the text and selections and
2. Open a comment box near each screen control for user to comment
3. It should also allow drawing arrows between controls
4. The user then enters entire problem or suggestion and then hits a “Submit” button
5. On the lines with the crash reporter, such a closure should post the form to the website of the software

There could be a bunch of marketing objections to such an idea:

• How to avoid duplications? We are already flooded with requests! [So I will throw the baby out with the bathwater.]
• What guarantees that such an implementation will result in more business? [So the ostrich algorithm is the best one.]
• None does that! [So, none needs it. The cart before the horse will run faster because it is already ahead.]
• Customer needs are so different, we will have conflicting suggestions. [I can prove it without facilitating.]
• It is just a UI! [ The best marketing strategy ignores the first word of what UI stands for.]
• It may yield only to a better horse carriage, not to a car. [This is a golden argument, provided your next product is going to make such a change.]
• It is useless! [So is the crash reporter.]

What do you think is the real reason?

Why doesn’t FOSS community adopt this approach universally? They NEED to improve usability. Can’t someone come up with libraries to achieve this?