search

cirosantilli / write-free-science-books-to-get-famous-website

MOVING to: https://ourbigbook.com Mission: live in a world where you can learn university-level mathematics, physics, chemistry, biology and engineering for free whenever you want from perfect open source books made for free by random people who want to get famous to get better paying jobs. 使命：生活在一个世界上，您可以随时随地免费获得完全随机的，希望成
https://ourbigbook.com
130 stars 18 forks source link

Some problems in the concepts #3

Open jedieudonne opened 5 years ago

jedieudonne commented 5 years ago

Hi, Your algorithm to rank "teacher" is interresting but it has some drawbacks.

1/ This algorithm is perfect for single questions answering. On stackoverflow, it's perfect. You give a good and wel explained answer, you get up ranked. And that's all. In this case, it's easy to know if the "teacher" was good : He answered the question or he didn't. But for a course, it's much harder. When I teach computer science my first course is about the print function. After one hour the student only knows how to write "hello world". So I'm a bad teacher. But in the course I talked about types of objects, which will be very usefull.... some lessons later.... And all my students say that this trick is part of my expertise in computer science teaching. But they say that, because they don't judge only after one lesson.

2/ People will rank what they like, not what is the most usefull for them.

In conclusion : If a teacher wants to be well ranked, he has to make people feel good, which is different of making people clever. I made you believe you intelligent so I'm well ranked. I asked you to make efforts to understand hard notions, I'm badly ranked.

My suggestion : keep you algorithm, he's usefull. But create a "expert council" to judge of the real pedagogical quality of a course. With those two indicators (expert ranking and user ranking) you'll get a very precise view of the course.

cirosantilli commented 5 years ago

Thanks for the feedback,

Yes, as you mention, those problems are too subjective to be solvable by machines, so I don't think much about them :-)

To me the main question is: is there an algorithm which will improve things, and which one is it.

Any algorithm can only be an approximation of "true value". PageRank managed to approximate "true value" well enough. I wonder if we can improve things significantly. Guess I'll only know when I try this out.

jedieudonne commented 5 years ago

I think your algorithm is great, but you can improve it with just this little addition

In parallel with user ranking, just include a expert ranking. This expert ranking would be delivered by people who can justify a real expertise in the domain covered by the course.

This way, you'll have two indications, user ranking will tell you how nice and how satifying a course is, and expert ranking will tell you how serious is the teacher, and how strong is the content.

And you don't have to modify your algorithm, you just have to apply it to two differents populations : users and experts.

cirosantilli commented 5 years ago

@jedieudonne OK, I'll keep that in mind.