This choice can be thought of as more driven by personality and preferences on how to think about a subject, than by a specific need to do one thing or another.
The question is not directly about personality theory, so I will just quickly reference the Big Five Personality traits (where this particular trait is hard to pick out) and Myers Briggs where it is clearly on the Sensing vs Intuition axis.
So my answer is that you need both approaches, they are both valid ways of dealing with building knowledge. They are also two equally valid ways that consumers of your world will approach understanding it - focussing on just one angle is making it harder for 50% of your audience to appreciate the work. Your task is not to choose one or the other approach, but to create a setup and team structure which helps all of you engage with the project.
This is not a trivial endeavour, because building separately top-down and bottom-up is not guaranteed to join up. However, you need each other. The big ideas need to support interesting details, not lead to blandness, or turn out to contradict themselves when thought through in detail. Conversely, complex details that don't fit into a theme or structure risk being a homeless mess of ideas.
The answer in my opinion is checking output and willingness to adjust towards the end goal. That means peer review, respecting each others' angle on how the world works. Where possible build new things that make the details possible, or things that explain the details in the context of the wider picture. This is just the improvisor's advice of "Yes, and . . ." re-worded. Where not possible, discuss suitable edits and revise.
I also suggest repeated rounds of brainstorming how things could and should fit together, as a team effort, as part of a routine for the project. Depending on how quickly the team is producing content for the world, this might be something you would do weekly, once per month, or maybe case-by-case as a peer review on draft content before the group agrees to accept the design and consequences of any piece of work as "canon" of the project world.
This approach works better when contributors can create drafts and plans of their work, presenting them to the group before committing hours of effort to polish them. It is much harder, and the producer less likely to be willing, to significantly modify a large almost-complete piece of content, as opposed to making the same adjustments to an initial draft.