Hard to tell if it will work a priori as efficacy will depend to a large extent on the nature of the Argonaute proteins and cofactors in your biological system (if small RNAs have been sequenced for your organism, this will give an effective shRNA size range). You might consider using slightly longer shRNAs with all the sequence similarity contained within the 5' of your shRNA guide strand. Most models for small RNA target recognition place considerable emphasis on the seed region, and nucleotides spanning the Argonaute cleavage site.

There are two questions here (at least): will the RNAi machinery be able to process such a short hairpin? I doubt it - enzymes usually require certain sterics to recognize and process their targets. However, you might extend your 14bp to make it longer - e.g. by adding some carefully selected scramble sequence and maybe by including several copies of this elongated sequence.

The second question is, whether a recognition of 14bp is enough to induce RNAi.

The shortest stretches of sequence which induced off target effects (i.e. they are active) were about 16bp, as far as I remember. These were part of regular siRNAs.

Hmm - appears to be a risky thing and hard to tell without trying.