This article:
Jon Fraenkel & Bernard Grofman: Does the Alternative Vote Foster Moderation in Ethnically Divided Societies?, Comparative Political Studies 39,5 (June 2006) 623-651
found that, contrary to IRV propaganda, switching to IRV voting unfortunately does not "ameliorate ethnic divisions" and encourage "moderation."
Their Abstract [note, what is called in the USA "instant-runoff voting" F&G call "alternative vote (AV)" as is common in Great Britain]: Using insights from social choice theory, particularly Black's concept of single-peaked preferences and Downs's median-voter model, the authors previously investigated the theoretical impact of the alternative vote system (AV) on the success of moderate parties in an ethnically bipolar society. Focusing on the simplest case, that with one moderate and one extremist ethnic party associated with each of the two ethnic groups, they found that for AV to necessarily yield outcomes that favor moderate parties there must be majority support for moderation and voter preferences that are single peaked with respect to the ethnic-conflict-defined dimension. Here, the authors test these assumptions with data from the 1999 and 2001 elections in Fiji, an ethnically bipolar society. They show that Fiji's objective of ameliorating ethnic divisions by the adoption of AV was not successful. In elections in 1999 and 2001, moderate parties would have fared better under a proportional representation system.
About the fact IRV favors 'extremists' & disfavors 'centrists'