I could concoct a few things, but I think it might be best
to use "tooth", a standard term in describing Arabic letters.

hhp

3 in Arabic = 1 + 3 in Latin?!
Rotate 3 90 CW, and align 1 to the left.

That's something Frutiger would do...
http://themicrofoundry.com/image/s_rome3-2.gif

hhp

I agree with tooth suggestion, that what I'd use myself. Also note that Arabic two can be written with a horizontal stroke with no teeth at all (This what children are taught in Egypt and how I write it by hand myself), but very few fonts implement it. This from KacstOne font.

Khaaled, that toothless 2 is a great idea, legibility-wise!

hhp

Indeed, AFAIK it was proposed by the Academy of the Arabic Language in Cairo specifically to cut the confusion between Arabic 2 and 3.

So now I'm dying to know: what else did they propose?

hhp

The Academy is mainly concerned about linguistic issues, but one of their interesting type related proposals was their simplified Arabic typecase where each letter has only two glyphs; one for initial and medial forms and one for isolated and final forms, with the main intent of decreasing time and cost of the printing process. Those that typecase were never used (expect in composing the booklet proposing it), it can be seen in later development's like Linotype Yakout.

There were also some recommendations about the positioning of diacritical dots and vowel marks to make it less ambiguous (calligraphers tend to position them freely on artistic basis, sometimes making text unreadable).

> Those that typecase were never used

why? publishers didn't like the idea?

May be, may be because it were too ugly as Yakout (which implements the same ideas) were picked by newspapers and remains in use even today. Also the largest Egyptian press by then, Bulaq, later simplified their typecase but in a much less drastic way (from 470 to 238 glyphs without counting vowel marks, compared to 135 glyphs in the Academy case including vowel marks).

By the way: aren't ٢ and ٣ ligatures of II and III ?