This is a non trivial problem, as demonstrated by this blog entry.
There are problems with all the obvious options:
Update: This is for educational software. The user will hear an audio Y/N question "Is the sky blue?" and select an appropriate answer. We can not expect the user to be able to read.
There is an ISO standard sign for "No", as stated in the blog post you mentioned. "Yes" is undefined, unfortunately.
Anyway, the signs are just (learned) convention. So my proposition is to borrow standard "No" and use "Yes" sign, which is encoded in different shape and color for easy recognition, see the picture.
To learn the convention you could perform a little training session for the patients before using the system:
Would it be possible to approach the problem in a different way? Instead of asking questions where the user decides just yes or no, make them state their choice by affirming text.
For example, instead of:
Does a pilot's license require the applicant to know the complete rulebook?
[Yes]
[No]
use:
A pilot's license requires the applicant to know
[the complete rulebook]
[the important parts of the rulebook]
Using yes or no symbols can be confusing, especially if the question includes negation (aka. "Do you really want to quit and not safe your work progress? Yes/No"). Reiterating the user's intent in a short phrase can be more understandable (i.e. "Save before quitting/Dismiss changes")
This sort of evades your question, but I think there is a good reason why there is no universally agreed upon sign for yes and no. These short binary answers always have meaning only in their context and not by themselves, thus creating a symbol for them is hard.
Considering your edited question my answer and examples are less applicable to your problem. Maybe one similar approach to solve the problem transferred to your situation would be to offer the users choices that still are more explicit than just a "yes/no" or attempted icon for those answers. For example your question "Is the sky blue?" could have two images as choices, one bright blue daylight scene and one orange-red sunset. This way you make the user repeat and affirm their choice.
I’d go with graphical animations of corresponding gestures.
In most cultures this would probably be a head shake and a nod:
Shaking to indicate "no" is widespread, and appears in a large number of diverse cultural and linguistic groups. Areas in which head shaking generally takes this meaning include Indian subcontinent, Middle East, Africa, Southeast Asia, Western Europe, and North America.
Nodding to indicate "yes" is widespread, and appears in a large number of diverse cultural and linguistic groups. Areas in which nodding generally takes this meaning include the Indian subcontinent (note that the head bobble also shows agreement there), Iran, Southeast Asia, Western Europe, Latin America and North America.
For cultures with different meanings for head shake/nod, you’ll have to adjust it accordingly. Of course this is only possible if you know the background of your visitors. If you expect visitors from many different cultures, you could start with an opening question like "Which gesture represents yes for you?" and let your visitors choose.
Not the best example, but you get the idea, right?
The dire restriction here is implicit, that a single pair of symbols must work for all cultures.
If you had several symbol pairs, and as part of the translation you could provide links to the locally appropriate pair of symbols then you might use the preferred choice for each culture.
OK, the i18n software does not allow for images translation (or does it?).
If not then have two hidden items like "yes_icon_URL" and "no_icon_URL" translated to the appropriate values, and use a small bit of js to set the button backgrounds in a localized fashion with the URLs contained there.
LOL to the ISO having an idea for 'No' but not for 'Yes'.
I'm in the UK, their 'No' looks to me like 'prohibited', not 'no'. Mind, I'm not in education and, as a driver, I see that sign on the roadside a lot, where it means just that.
You've given some context with referencing educational software but not said what age group. I work in that realm and with primary school age kids. I'd definitely not go with green squares, although the colour is right, obviously - for those without colourblindness, at least - the shape means nothing.
I'd go with ticks and crosses first and then vote for the smiliy face.
