It seems today's theme in the aid blogosphere (Aid on the Edge, Duncan Green) is Positive Deviance, so I thought I'd jump on the band-wagon because I love the concept. Positive deviance is the epitome of solutions emerging from practice.
Basically, in every situation there will be an average but there will also be positive and negative 'outliers'. So, in HIV prevention, 12 villages may have very similar situations in terms of socio-economic status, average educational levels, unemployment, poverty, etc. So these villages are roughly equivalent. Now, say HIV, STI and pregnancy figures are compared for the 12 villages and it turns out that 3 of the villages have much higher rates of HIV, STI and teen pregnancy and one village has much lower rates than the others. Most approaches to development planning will work with the average and invest in a tried-and-tested BCC, peer-educator and/or health-service upgrade approach for all 12 villages. The positive deviance approach says that (potentially in conjunction with standard, outside-tested methods), you should go to the positive outlier village (with lower infection and teen pregnancy rates) and try and find out what they're doing differently.
This has two impacts. The first is that, by investigating and often publicising that one village is doing something right, something that the outsiders (with all their power) want to learn from, you reinforce the positive deviant behaviour. This is obviously good for the village because it is likely to strengthen whatever they're doing right, thereby increasing the positive deviance effect. Secondly, if you're lucky, you may discover a simple method or approach or 'thing' that this village is doing, that can be clearly linked to the outcome and may even be sufficiently transferable to apply in the other villages, thereby protecting an even greater number of people.
Without rejecting the value of outside knowledge or the usefulness of international experience, positive deviance makes it possible to identify, support and sometimes even replicate locally-developed (and therefore generally more relevant) solutions. This can sometimes be as simple as getting the mother whose child is not malnourished to teach other mothers how to cook.
Solutions are grounded in reality, rather than being imposed from a head-office miles and miles away. Local people are recognised for what they're getting right, making them far more equal partners than when they are simply, patronisingly, asked to approve of what the aid workers tell them is the best option. Also, specific individuals, who are doing things better, are acknowledged and given more support, which, eventually, creates the possibility of more general behaviour change (provided possible backlash is carefully managed and aid/development workers are willing to get over their pride and give the credit to the local person).