We're having a little discussion in the Teachers' Lounge (a site-wide chatroom for ♦ moderators).

Sometimes, we need to destroy accounts (e.g. in the case of spammers). This is an operation which involves going to the user's profile and a couple of clicks in the following dialogs (not unlike flagging a post or voting to close it). Most dialogs on Stack Exchange have keyboard shortcuts enabled, which (once you get used to them) are a lot faster than clicking.

The keyboard shortcuts for destroying an account are almost working, but one of the moderators argued that one shouldn't use keyboard shortcuts for a dangerous, irreversible process like this; it's just too easy to make mistakes, destroying a user even if it wasn't your intention. For those of you who are worried: destroying a user with keyboard shortcuts still takes five or six keystrokes; that's unlikely to happen by accident.

This discussion lead to my (generalized) question. Has there been any research on the following:

Are users more likely to make mistakes with functionality which is easier to access?

(Of course, if it's easier to access, it will probably be used more often, leading to more mistakes. I'm interested in the relative amount.)

Personal experience with this could be helpful as well, but please refrain from misusing anecdotal evidence (of the "I once made a mistake because of ..." type).

  • 41
    If "destroy user" is still irreversible like I remember it being, then a keyboard shortcut is absolutely a bad idea. This should take at least 2 clicks to prevent misfires. – Cody Gray Jul 26 '17 at 11:24
  • 1
    The two answers below are useful (thanks, I upvoted them), but I'm also interested (if it exists at all) in some kind of graph relating the # of clicks (or a similar metric) and the error rate. – Glorfindel Jul 26 '17 at 11:28
  • 8
    If this feature is crucial but kind-of-frequently-used, it may be really useful to have something like a Trash Bin for users - moderators Destroy a User, and it is kept in the Purgatory for a month so it can be Un-destroyed if a mistake has been done. – mgarciaisaia Jul 26 '17 at 19:25
  • 7
    Note that only the last keystroke needs to be wrong to delete a user inadvertently. The previous four or five are also used for other actions you can take on that user, so it doesn't need multiple mistakes to get it wrong, just one. – terdon Jul 27 '17 at 10:07
  • 2
    Some already mentioned that the keystrokes may overlap with other shortcuts. Keep in minds that there may also be other things that could overlap with the keystrokes (especially if you would have 2 windows in parallel). For instance anyone could take the name mm4a by coincidence or on purpose to illicit erroneous deletes. (Not to mention accidental occurrances of things like mma or mm4) – Dennis Jaheruddin Jul 28 '17 at 7:34
up vote 52 down vote accepted

Alan Cooper uses an ejector seat analogy in About Face which is pertinent here:

Just as a jet fighter needs an ejector seat lever, complex desktop applications need configuration facilities. Applications must have ejector seat levers so that users can occasionally dramatically (sometimes irreversibly) alter the application's function, behaviour, or content.

The one thing which must never happen is accidental deployment of the ejector seat. The interface design must ensure that the user can never inadvertently fire the ejector seat when all he wants to do is make a minor adjustment to the application.


If you need to trigger five shortcuts sequentially to trigger the action, and a combination of those shortcuts does not perform any other action, the ejector seat is very well hidden.

However, as Terdon suggests, in the OPs case the first four shortcuts in the sequence can be used to perform many different actions (mostly navigating menus).

You only have to get the final shortcut in the sequence wrong to potentially activate the ejector seat.

  • 1
    I like the analogy and agree, but I believe Glorfindel wants to know if there are actually more mistakes made if access to a functionality is easier. – Nick Groeneveld Jul 26 '17 at 9:48
  • 5
    About Face is research: it draws from a vast number of studies, and itself is cited in a significant number of studies :) However, if by research you mean an empirical study, the OP is unlikely to find it in a generalised manner as the subject matter would be much to broad. But I'm not sure the OP is looking for an empirical study? – Joel Tebbett Jul 26 '17 at 11:05
  • 23
    The plane is crashing in 2 sec. Please confirm ejection by solving a Rubik's Cube. – Guillaume Jul 26 '17 at 13:56
  • 5
    @Guillaume Ahh, but in that case you're not changing the position of the ejector seat lever, you're changing the operation of the lever altogether. It's possible to make it harder to accidentally trigger an action without making the action itself significantly harder to perform. As with most UX issues, there's a cost/benefit compromise to be made. – Joel Tebbett Jul 26 '17 at 14:19
  • 3
    It's basically like navigating a menu. Each shortcut opens a different submenu, so the preceding shortcuts can also be used for different actions. Only the last one selects the delete action. – terdon Jul 27 '17 at 10:17

I would say that instead of being a "mistake", in this case, it´s a "slip".

According to Nielsen Norman Group:

Mistakes are made when users have goals that are inappropriate for the current problem or task; even if they take the right steps to complete their goals, the steps will result in an error.

Slips occur when users intend to perform one action, but end up doing another (often similar) action.

Now, according to the book "Experimental Slips and Human Error: Exploring the Architecture of Volition":

Very simple tasks mean more opportunities to err because they can be accomplished more quickly. (...) Simple tasks tend to yield a large percentage of slips, whereas complex tasks result in more errors of intention, planning, and judgement.

So, yes, users more likely tend to make mistakes (slips?) with functionality which is easier to access.

A way to prevent slips is to allow the user to "undo" its action

  • 11
    Agreed. Instead of actually deleting the user account in the database system, it should rather be flagged as deleted. All posts made by that user can be hidden based on that flag, and the record eventually deleted permanently after a grace period of e.g. 7 days (before that, moderators would be allowed to reactivate the account and undo their slip). – CoDEmanX Jul 26 '17 at 14:06
  • 4
    Too be pedanticly precise, undoing is a way of mitigating a slip that occurred, not preventing slips from occurring. – The Nate Jul 27 '17 at 7:51
  • 1
    @TheNate agree that it´s a way of mitigating a slip that occurred but I do believe that slips or other human error can't be 100% prevented when there is a mental process by the user. The only way to prevent this is by using a mechanized process instead of a human – Joao Carvalho Jul 27 '17 at 8:01
  • Understand that my comment was entirely linguistic. I agree with your logic. – The Nate Jul 27 '17 at 8:04

Similar to Peter's answer, I would direct your attention to Amazon one-click purchasing. Doing a quick google for "accidental amazon one-click purchase," the first result you'll see is a forum on Amazon of users very displeased with the "ease" of one-click purchasing due to the users accidentally accessing this function.

This caused enough of a problem at the time that Amazon then reverted to two-click purchasing. I believe this is a prime example of mistakes made from ease of access to technology. Typically when you want to see the result of something, you look to those who will lose/make money from it either going well or going wrong.

You can actually get an estimate on human error by doing a Technique for Human Error-Rate Prediction (THERP) you can read more about that here THERP Wiki

This would allow you to conduct your own research and see how much of an impact it will have.

  • 5
    Amazon still has one-click purchases. They just make it very simple to cancel an order if you do it before it is dispatched. So I believe the solution for this is to make deleting users accounts very simple with keboard shortcuts, but give the possibility to revert it in a less easy way during a few minutes. – ecc Jul 27 '17 at 10:29
  • @ecc You are correct in that is what they do now, the situation I mentioned was from 2011, when it wasn't as easy to cancel. They have since changed their system to adapt. I like your solution though, make the shortcut, but allow a grace period to reverse it. – CumminUp07 Jul 27 '17 at 13:07
  • Indeed, Amazon provides a couple of escape paths to make up for the ease of one-click purchasing. After clicking, you get an order confirmation page, so that you know that you've placed an order. A short time later, you receive an order confirmation email, and can click through to the website to cancel. Later, you can refuse delivery of the package when it arrives, and you'll eventually get a refund. Failing that, you can return it (though you might have to pay return shipping unless you talk to customer service about it). The difficulty of these steps escalates progressively. – Zach Lipton Jul 30 '17 at 5:41

I would say absolutely. There is the Ctrl-Enter keystroke in Outlook to send an email immediately, and I love it. Very rarely do I ever inadvertently send an email. However, when I'm having a conversation in Skype, I'll occasionally use the same keystroke and start a call with that person instead of send my message.

Thankfully, it's only every other month or so, but when I'm rushed or stressed from what I'm working on, I'm far more likely to forget what app I'm in and use the wrong keystroke.

Similarly are websites and apps which utilize Shift-Enter in a textarea to go to a new line instead of submitting the form. Then there are other sites which capture the Enter keystroke within the textarea to add a new line, and the Shift-Enter does nothing. I get those sites confused quite often.

One would hope that admin duties, especially nuking a profile, aren't done while under the same stresses as having been fighting with a section of code for 3 hours, but accidents do happen. If the resulting action is completely irreversible, I would at least recommend a confirmation dialog with the keystroke.

  • 5
    The problem about confirmation dialogs is that we're all trained to click them away before reading them, because so many of them are pointless. The strongest protection for such irreversible and potentially harmful actions is to require the user to type "DELETE" into a textbox. – Peter Jul 26 '17 at 15:40
  • @Peter Good thinking! – krillgar Jul 26 '17 at 15:41
  • @Peter World of Warcraft does this when you try to discard a rare or epic item. Normally you sell that stuff to vendors, but a few can't be sold. And some of those are both common and useless (looking at you, AQ40). You end up typing "DELETE enter" reflexively and without thinking. There goes that. – Harper Jul 28 '17 at 1:06
  • By the way, I called my manager instead of sending the message I typed yesterday afternoon. :( – krillgar Jul 28 '17 at 10:49

I offer some original research:

Google these words: Internet Explorer Backspace

The titles of the top 10 search results are:

  • How do I disable the Backspace KEY from functioning as the Back ...
  • Disable Backspace Key in Your Browser for Peace of Mind
  • javascript - Prevent backspace button from navigating back in ...
  • javascript - How to disable the Backspace button in Internet explorer ...
  • How to disable the Backspace button in Internet explorer using asp.net ...
  • Disabling Windows Backspace = back option - Ask for Help ...
  • keyboard shortcuts - Why does backspace go back a page? This ...
  • Disable Backspace IE Navigation - Technology Group of the North, LLC
  • Disabling backspace in Internet Explorer | PC Review
  • Desparately need to disable Backspace Key as Back Button - Google ...

As you can see, 10 out of 10 hits want to disable this functionality, which is easy to access accidentally, sometimes for catastrophic effect.


Since you also asked for personal experience, if you've been around during the time when backspace was the standard default shortcut for going back in most browsers (2016?), you will have a personal experience with the backspace shortcut that was less than satisfying.

You will also remember that you didn't have nearly as many unsatisfying experiences with the back button, despite using it far more often than the shortcut.

  • 1
    I must say, I find myself using my mouse to click "back" - or simply not going back, since backspace no longer go backs. Old habits. I guess what I'm saying is if you change a behavior to prevent it, you might also prevent people who like using that behavior. – James Haug Jul 26 '17 at 21:51
  • 11
    10 out of 10 hits want to disable this functionality, because it exists and the person who likes that shortcut has no need to google it. Now try google 'safari go back delete' instead... – Coxy Jul 27 '17 at 0:16
  • 5
    @Coxy has identified the confirmation bias upon which this answer is based – cat Jul 27 '17 at 2:34
  • 3
    @Cozy You misunderstood. This isn't about confirmation bias at all. The example is about people wanting to get rid of functionality that is easy to access accidentally. The 10 results prove they exist in relevant numbers. If the shortcut wasn't easy to access accidentally, people wouldn't want to get rid of it. Neither the question nor this answer discuss if backspace is a great shortcut. Confirmation bias would be if I used these 10 hits as proof that everyone wants to get rid of backspace as a shortcut that deletes all user input on a form without confirmation or option to undo. – Peter Jul 27 '17 at 9:17
  • 1
    @Harper Yet for a small percentage of users, it was desired behavior. I could see it being an option (possibly off by default) rather than removing it entirely - when Chrome removed it, I was completely lost for a few weeks until I found a plugin to restore it. – GalacticCowboy Jul 28 '17 at 13:38

Mobile devices are the ultimate "ease of access causing mistakes"

You know. Most apps strive to have every touchable area do something. And it runs on a gadget which strives to provide touch surface almost edge-to-edge on the narrow side, and the wide side is too wide to grip. (thank God for iPad's 3/4" margins.)

If that wasn't enough of a recipe for disaster, the touch screens are super bad at distinguishing a "tap" from a "swipe" (at least for me) and do the wrong one at least 1/3 of the time. This results in lots and lots and lots of backing out to where it becomes muscle memory.

Min-maxing for "max effect from minimum user input" is terrible.

It should be just a little bit hard to do things.

You can probably figure out where you've screwed up just by looking at where "undo" or "back out" gets used a lot.

  • In the mobile case specifically, I'd argue that if the "do something" button is on the right hand side of the screen, then the "confirm" button should be on the left of the confirmation requester (be it a popup or a screen of its own). Thus, it's still two presses, but you're more likely to cancel the confirmation than to enact it if you're doing it by accident. – Ralph Bolton Jul 31 '17 at 9:49

A rather obvious point (but so far unmentioned) is that the last click or key stroke should be physically away from potential other functions.

If you have multiple clickable buttons next to each other, the chance of a slight miss with serious consequences is much higher. Simply moving high-impact functions a safe distance away should result in a lot less chance for an accidental click.
The same applies for keyboard shortcuts, the letter or key for the function with serious consequences should not be right next to another key which applies at this moment, or applies to the overall program or OS.

Of course, the best option is to make any function reversible, if only for a minute; but that is not always possible.

The classic example for me is CTRL+W - which in the windows version of thunderbird would be used for writing a message.

In linux, it would close the window. This lost me some carefully worded message, and it could not be undone.

I do agree with krillgar that the Ctrl+Enter shortcut does not suffer from this problem.

The ejector seat lever is not easy to access.

Only specially trained and qualified users are allowed to get into a jet fighter airplane.

Likelihood of errors vs. magnitude of errors

In one project, a "shortcut" function (available only to specially trained and qualified users) did not increase the likelihood of an error, but the magnitude, by a factor of at least 100.

The opposite effect

Example: a user needs to alter the data or behavior of a web application via SQL injection because the standard user interface provides no efficient option to perform the required actions. Chances are pretty good that the user will actually kill the web application instead of reaching the goal.

Your Answer

 
discard

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.