Forbidding products causing those ailments (or using some more indirect similar way, for example over-taxing it) is counterproductive as 'helped' people then search for other ways, and their hostility towards the "helper" grows.
This illustrates effects of one of the strongest imposed ways, but is IMHO true for any imposed action.
Trying to honestly cooperate with the 'helped', instead of adopting a 'solution' defined by specialists in order to solve patent 'bad effects' then imposing it, leads to find and tackle at least some roots of the problem.
Making an argument for such an holistic approach seems possible to me because reductionist approaches aren't really efficient in the long term when applied to human beings.
Using abstraction in order to solve locally (<=> focusing on one problem at hand and neglecting the rest, even other parts of the life of its 'victims') leads to, later, discovering that the applied local solution isn't adequate, patching it, then having to patch the imperfect patch... (rinse and repeat).
This approach is popular because it makes wonders in a 'near-model' context (for example in order to develop software or even to build a bridge), but it seems less convincing to me when applied to human beings (especially when some make a living because the problem exists => they may not really want to solve it).
This is all very abstract; I am unable to identify one specific recommendation here as to how to reduce acquired diabetes and obesity in young people.
This discussion has drifted a long way from your original claim, "IMHO caring is giving what people want, not trying to do what you declare good for them." In the case of diabetes and obesity, patterns of consumption are formed in childhood. My wife knows a mother who took the position you are advocating here literally, and now that her children are in their late teens, it is resulting in all sorts of problems for both them and her.
For you, me, and everyone else, no matter what ideological position one holds, there are a lot of problems having no easy solutions.
The present thread was about the Gates Foundation (and similar institutions primarily targeting whole populations, that is to say adults and, through them, their children), and I answered 'give them what they want', meaning in this context 'when caring for adult people, give them what they explicitly ask for (not what you prefer)' and also (as such a Foundation has to aim for the greater good): 'if and only if you objectively think this will be good for them and not bad for others'.
Youngsters aren't mature, therefore we cannot merely give them what they want. We have to educate them, which is maintaining them in an adequate context while guiding them towards finding what they want and will be good for them and not bad for others. In my opinion our contemporary society very rarely offers such an adequate context. Hoarding kids in overpopulated classrooms where they sit for hours with adults having few (if any) personal interconnection with them is apparently the most efficient way... and (IMHO) a recipe for disaster (diabetes and obesity getting out of control being part of it).
I didn't wrote that my approach is simple and explicitly stated, above, ((that)) 'It is, however, more difficult' than the usual 'select a cure for the symptom and impose it'.
This illustrates effects of one of the strongest imposed ways, but is IMHO true for any imposed action.
Trying to honestly cooperate with the 'helped', instead of adopting a 'solution' defined by specialists in order to solve patent 'bad effects' then imposing it, leads to find and tackle at least some roots of the problem.
Making an argument for such an holistic approach seems possible to me because reductionist approaches aren't really efficient in the long term when applied to human beings.
Using abstraction in order to solve locally (<=> focusing on one problem at hand and neglecting the rest, even other parts of the life of its 'victims') leads to, later, discovering that the applied local solution isn't adequate, patching it, then having to patch the imperfect patch... (rinse and repeat).
This approach is popular because it makes wonders in a 'near-model' context (for example in order to develop software or even to build a bridge), but it seems less convincing to me when applied to human beings (especially when some make a living because the problem exists => they may not really want to solve it).