The problem with the way you state the issue is that you focus on the issue of spoken or written language first. But that is not really the problem. The question of natural vs. 'forced' learning applies to learning just the spoken language (or even just the written language).
This is a well known and well debated distinction in the SLA (Second Language Acquisition) literature. I have not seen the 'forced learning' term but people do talk about 'natural learning'. One of the best known (and influential in the 1980s) was Stephen Krashen's 'Natural Learning Approach' which started off with exactly the same assumption - we should learn second languages more like the way children learn their first language.
Building on the well-recognised 'acquisition' vs. 'learning' distinction, Krashen introduced a so-called monitor model that illustrated how learned material can slow down 'natural communication' and hinder acquisition. He also talked about the affective filter which in some ways could be labeled as 'forced learning' getting in the way of 'natural learning'.
Krashen's ideas were not new but built on foundations going back at least a hundred years. There are many other approaches that try to lower the affective filter and replace the monitor with more natural input going from 'Total Physical Response' to 'Suggestopedia'.
However, this approach did not really stand the test of time as singular catch all method. It made a lot of sense but the research to support it is sparse. In short, adults have many very useful cognitive strategies such as memorisation, system building, impulse control, metacognition, ability to read and take notes that can speed up their initial acquisition greatly when compared to children in a similar situation (e.g. attending a class twice a week). So, if they were to give up on those, they would forego a great advantage. However, you're still left with the problem of how they can transition from knowledge of the language to a more naturalistic level of proficiency. Here's where many of Krashen's (and others') ideas have taken better hold. I personally find his 'comprehensible input' hypothesis quite a useful way of thinking about success in learning. And many of the internet polyglots (probably knowingly) mirror many of Krashen's suggestions when describing their ultimate success.
However, the complete picture is even more complicated. You need to take into account learning strategies, learning styles, the individual's situation, etc. Which is why the current approach to SLA practice is one of eclecticism (for instance, when people speak about the 'communicative approach', it is an amalgam of many of the methods that came before).