You should actively not use reason in situations where your ability to reliably reason about the accuracy of your reasoning is known to be deficient.
Our ability to use reason to make a convincing case (to ourselves or others) is insanely higher than our ability to actually make a correct case.
In some areas, we have sufficiently simplified the world (or developed tools that handle the complexity) that reason is a great tool to interact with it.
Look at the cases you judge as being reasonable to use reason:
- designing a marketing project
- writing a scientific article
- drafting an architectural plan for a house
- preparing a shopping list
- planning their holidays
Designing a marketing project, we have entire fields of expertise and social systems -- economies -- designed around making the selling of goods and marketing them as easy as we can.
For scientific articles, the goal of science has been to develop a way to reason about the world. Problems that are not suitable to reason about, or that you haven't succeeded in working out how to reason about, aren't good scientific articles; they won't be useful to the consumers of the article.
Engineering and Architecture is also a huge tower human effort into making things simple and effective. We have building codes and patterns and materials and designs that are insanely easier to reason about than the general problem of "I want shelter that is good". You can know the material properties of a steel beam before you even start building your house.
Shopping lists rely on an economy that produces many copies of the same kind of good and stores them in an easy to find and consistent spot for a predictable amount of economic purchasing points. Again, a huge amount of effort, the collective work of billions of humans over 100s of thousands of years, making the problem really really simple.
Even modern holidays rely on huge amounts of tech, from calendars to transportation networks and stable currencies and the like.
Every one of your examples has a tower of human civilization's efforts making the problem amenable to reason. Next to none of them have huge incentives on the other side making people want to avoid making it amenable.
Now, look at your second list:
- making friends
- dating boyfriends or girlfriends
- judging the action of other people
- the veracity of outlet news
- speaking up in public
- arguing with subways travellers
For a bunch of these this is matters of interpersonal trust. Trust is a war; if I can figure out how to make an arbitrary person Trust me using simple and cheap techniques, I can cause nearly unlimited harm to them and use that Trust to take resources from them.
We have structured society in such a way that Trust is insanely easier than it would be without society. We have signifiers of status and identity and community that can be used to make certain kinds of Trust-attacks less likely or more expensive, we have state monopoly on force attempting to make it less dangerous, etc.
But the incentive to exploit Trust means that people are working out ways around the system with lots of incentive. So we can't rely on reason, because we have failed to simplify the system, because people have incentives to keep it from being reliable and easy.
News -- politics -- is a similar Trust game.
The remaining bits looks like social mores and acceptable behavior in public. Here you are mainly signalling you understand the rules of society and abide by them; the actual communication doesn't seem very relevant.
And even the earlier ones - the ones where we have simplified things - remain a battle. Marketing can be viewed as a war to sell useless stuff to marks, scientific articles have to convince reviewers and readers that you should have academic social standing, houses have to look high quality to sell to buyers and advertise their social status, people selling goods would love to sell you worthless stuff for high prices, and the like.
In many cases, we end up with regulation. Grocery stores can't sell rotten food, you can't build a house breaking code, etc. These again attempt to simplify the transaction, and make reason more effective and easier.
In short, humans are great at reason in situations we make reason easy to do. In more complex situations, humans are often overconfident at their ability to reason; they have been trained on the easy situations where reason works well, and project it into harder domains. In those harder domains, using the same reasoning techniques that work fine in the simple domains results in rather poor results.