Boris Smus

interaction engineering

Interactive touch laptop experiments

Largely because of the plummeting price and thickness of touch screens, these devices are increasingly ubiquitous. One of the latest trends is touch screen laptops, spearheaded by Surface devices and the recently announced Chromebook Pixel. In this post I'll dive into some experiements around this new form factor. The main goal is to try to convince myself that this form factor makes sense for reasons other than economic ones.

In exploring the interaction design angle of these new devices, I came across a couple of what I think are a couple of interesting ideas that I'd like to share with you: responsive input and simultaneous interactions using both mouse/trackpad and touchscreen. I wrote some demos that illustrate these ideas. A touchscreen laptop is required for these demos to work properly.

Because you probably don't have a touch screen laptop, I recorded a rough video showing some of these interactions:

Hopefully this gives you a better sense of what I mean by responsive input and simultaneous touch and mouse interactions.

Responsive input

The touch laptop class of device has a two main interaction modes:

  1. As a regular laptop with trackpad (or external mouse) and keyboard.
  2. As a touch tablet with a keyboard.

These two interaction modes differ fundamentally in many ways. The following are some examples of these differences:

  • Touch has no hover state.
  • Touch is less precise than mouse and requires bigger targets.
  • Touch requires that you are closer to the screen.

Chrome Pixel

Ideally, you want to provide optimal experiences for both cases. For the mouse case, this means taking advantage of hover states and a finer pointer. For the touch case, this means ensuring that touch targets are big enough to be tapped, not relying on hover at all.

So I explored a user interface concept that adapts touch laptop interfaces to the user's current input mode. The tricky bit is detecting the user's current input mode. Several adaptation options are possible:

  1. Immediately transform to the mouse-style UI as soon as the input mode changes (simplest, but can cause transitions to fire too rapidly between the two modes, which may be jarring).
  2. Transform only after some period of not using the other input mode (eg. go to touch mode only if the user is actively using touch, and not touching the mouse at all).
  3. Transform based on some external criteria, like whether or not the screen is docked to a mouse, or based on input from sensors other than mouse/touchscreen.

The first approach is problematic in that your first touch transforms the page. If this transformation causes your target to move away from it's initial position, you will miss it entirely. This can be mitigated by having intelligent resizing which does not affect anything directly under the touch point, but may result in a lopsidedly zoomed interface.

The second approach is problematic since the mode switching will happen automatically after some period of inactivity, which may be jarring. The last approach is either obvious (eg. mouse removed), or an area of research (eg. predicting when the user will touch based on camera).

I wrote a demo of auto scaling in response to input type. If you use the mouse, click targets will decrease in size. If you use your finger, touch targets increase in size. (Of course, this will only work on a touchscreen laptop).

Simultaneous touchscreen + mouse/trackpad interactions

In the above section, I described an automatic way to switch between touch and mouse mode However, there is a middle ground between the two: multimodal interactions that involve both touchscreen and mouse/trackpad. Simultaneous bimodal interaction is already common. For example, using mouse and keyboard simultaneously makes a very efficient interface for FPS games, with the movement via the WASD keys, and mouse-look.

One experiment involves using the mouse or trackpad as a navigation device and using the touch screen as a way to input positional data. This is illustrated through Google maps. You pan and zoom the map using mouse events, and place markers on the map using the touch screen. Try out this demonstration of mouse-to-map and touch-to-mark (again, this requires a touchscreen laptop).

Another experiment involves manipulating geometric objects on the screen. The idea here was to use the touch screen to select objects, and use the trackpad/mouse as way of manipulating the selected object. In this demo, you can manipulate the object in a number of ways:

  1. Move it by simply dragging it around on the screen with touch.
  2. Rotate by selecting the object on the touchscreen, and then performing a mousemove (either by moving a mouse or dragging one finger on a trackpad). The rotation happens around the point where you touched the object, which acts as a fulcrum.
  3. Scale it in the same fashion as rotation (selecting object and transformation origin with the touchscreen), except with a two-finger drag on the trackpad, or using the mousewheel if a mouse is attached.

With no selection, the canvas itself can be zoomed and panned with the mouse/trackpad directly. Try out this multimodal transform demo (requires touchscreen laptop).

Missing pieces

Like any brave new world, the one of multimodal input has its own set of challenges.

It's currently impossible to distinguish a touch laptop from any other touch screen. Notably, this means that you should never assume that touch support implies no mouse support. In practice, make sure that you always bind to mouse events. If you also have touch event handlers, just use event.preventDefault() there to ensure that you aren't handling one event in multiple handlers. If you're interested in this, follow the discussion at

As a generalization of the above, there is currently no way to determine which kinds of input are available in the browser. A fully fledged Input Availability API might seem like overkill, but there are already some cases beyond touch laptops that are relevant. For example, detecting the presence of a physical keyboard would be very useful. Further, detecting hardware features like an attached camera and microphone could fall into the same bucket rather than relying on exception handling from APIs like getUserMedia. Lastly, having such an API would allow websites to react dynamically to changes in input (eg. a tablet gets docked to a physical keyboard, or a mouse is attached).

The final missing piece is that dealing with two different event models (mouse and touch) is definitely clunky. I have already written extensively about pointer events and a pointer event polyfill. In this particular case, pointer events would be great, because although they provide a consolidated model for input, it's very easy and natural to distinguish between the two modalities.

These experiments are all available on github.

Your turn!

Do you have thoughts or demos around new types of interactions using touch laptops? Please share them below.