Re: [w3ctag/design-reviews] The `interesttarget` attribute (Issue #1058)

xiaochengh left a comment (w3ctag/design-reviews#1058)

**Note: this comment is NOT a TAG consensus, but my personal thoughts only.**

TL;DR: To me, `InterestEvent` is the most interesting part of this proposal, as it directly defines a user intent that is not bound to any specific action. There might be great potential following this path forward. How this event should be triggered might be an orthogonal question.

---

### Status Quo

In all current UI designs, conversions between user intents and actions are involved:
1. User has an intent
2. Facing a specific UI, user must convert the intent into specific actions. The conversion is based on the user's knowledge of common UI design patterns and guesstimates, and sometimes fails.
3. The UI receives a sequence of events
4. The UI reconstructs the user's intent from the sequence of events. The reconstruction is based on an action --> intent mapping programmed into the UI (either by UA or by authors).
5. Based on the reconstructed intent, certain business logic is invoked, and a response is presented to the user.

For this workflow to success, the user must understand the UI design language to perform the correct actions.

### With "Intent Events"

`InterestEvent` is the first event that does not define any action, but a user intent directly. Imagine if we can define a whole set of such "Intent Events", then there are a lot of different things we can do. I can imagine the following, ranked by how far we go from the status quo:

0. User still needs to convert their intent into actions, and UI still needs to reconstruct the user intent. However, the reconstruction dispatches an `IntentEvent` to trigger subsequent handling. So far it's still not very different from the status quo, just an additional layer of abstraction.
1. The triggering of an intent and the handling of an intent can be completely separated. For example, developer/UA may design different ways to trigger an `IntentEvent` with different input devices (hover with mouse, long press with touch, certain keywords with a speech input, staring with a serious face using eye tracking + facial recognition, ...)
   - Different apps may even experiment different ways to trigger a certain `IntentEvent` and compete. UA may adopt the survivor as the standard way.
2. New apps that are completely driven by `IntentEvents` are developed
3. AT might be improved, as they can trigger Intent Events directly, and drive the app on the same API surface.
4. Since there's a standard set of Intent Events, the user may even customize their own way to trigger each intent in user preferences, and no longer require any guesstimation in the UI design language. 
5. User privacy is improved since they can no longer be fingerprinted by their exact actions. In some sense, users are more distinct yet less trackable.

### Path to AI user agents?

Nowadays, it's hard to avoid AI when discussing technologies: What will AI agents be like? How is it different from a conventional UA? Will Web still be relevant? ...

The most convincing theory to me is: with an AI agent, user can directly express intents without the need to convert it into specific actions.

Intent Events seem to be a solid first step towards such future. User can use AI to trigger Intent Events and drive the app. Furthermore, any existing app developed with Intent Events can be easily adapted to an AI agent, so we can solve the problem of relevance.



-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3ctag/design-reviews/issues/1058#issuecomment-2883570301
You are receiving this because you are subscribed to this thread.

Message ID: <w3ctag/design-reviews/issues/1058/2883570301@github.com>