The study uses an agent-based computational model to test the hypothesis that contact between two dialects that are conservative and innovative as far as a sound change is concerned produces an asymmetric shift of the conservative speakers towards the innovative ones. The computational model, which is initialized with speech data from real speakers, is tested for the first time on a morpho-phonological sound change by which cues to inflectional morphology are being transferred from a suffix to a stem vowel in two Italo-Romance dialects of Southern Italy. The results based on quantifying the extent of diphthongization and the emergence of categorical contrasts marking morphological inflection provide some support for the proposed asymmetric shift in dialect contact. The analysis is more generally consistent with feedback models of sound change in which production is stochastically updated as a consequence of memorizing speech signals in speech perception.