There are two main ways of dividing discussion in a group so that people only see what they want to see and both have advantages and disadvantages.

Tags are the most common method used. They have the following advantages, mostly related to flexibility:

Anyone can invent a new tag on the spot and they can organically gain usage
Posts can be given multiple tags. This allows you to filter posts on multiple dimensions (ie. topic and type of post), as well as having both broad and specific tags. In a well designed system, this provides maximal control on what posts you want to filter in/filter out.

Sub-Groups are less common. They have the following advantages, mostly related to the ability to create norms:

As demonstrated by Reddit, having different sub-groups allows you to set rules that work well for the community. For example, Ask Science can have a norm of encouraging peer-reviewed references, Ask Me Anything can delete posts which are abandoned and Change My View can ensure that top-level posts actually try and change your view. While this would be theoretically possible with tags, this is complicated by the fact that a) posts can have multiple tags so it is unclear how these rules would interact b) it is easier to add a tag without ever understanding that there are rules you are supposed to follow, than it is to accidentally post in a group
Even if you can set rules for tags, it is much harder for implicit norms to form for tags than for groups. The fact that a post can have multiple tags really torpedos the formation of implicit norms, as it prevents there from being a clearly defined space
Sub-groups tend to have much more a sense of community, indeed they are explicitly a community, vs. any implicit community that occurs via posting on the same tags
Sub-groups tend to result in less duplication, while there may be five different tags which are all effectively synonyms of each other

My position is that the value of people forming separate sub-groups should not be underestimated. Think, for example, how helpful it would have been in the early days for rationalists interested in Effective Altruism to be able to split into their own group. Much of the discussion, ie. of specific charities, would have been of no relevance to most of the LW community, but of vital importance to that particular community. This brings us to another distinction. A parenting tag, for example encourages discussion at the intersection of Rationality and Parenting, while a parenting rationalist group encourages discussion which Rationalist parents might be interested in. These are not the same: the second is broader as many of the discussions that Rationalist parents might want to have would actually just be generic parenting discussions, but with people who have a rationalist world view. I believe that this option leads to groups which offer more value to their members, (but with the cost being that the top-voted posts can't be fed into the main stream as many of these posts won't be directly related to rationalism).

Which method should be preferred on LW2.0?