Etica.AI is concerned about long-term social & ethical implications of artificial intelligence on Africa/Latin America. Do it via grassroots acts in short-term
There are several ways to transfer knowledge to humans. Efficient ways of doing this are very important, even in a machine-intensive environment. "Learning by Teaching" is one way to do this that has advantages in the quality of learning at the cost of more time to do something.
Thinking about the future, it pays to prepare for technologies that will come instead of taking traditional methods of teaching in schools of our time. See also Zone of proximal development from Lev Vygotsky, that may explain why, even with vast knowledge on the internet, people do not learn without some kind of direction. Dale's Cone of Experience also have some, not fully, impact on retention level of knowledge.
In professional education, learning by teaching (German: Lernen durch Lehren, short LdL) designates currently the method by Jean-Pol Martin that allows pupils and students to prepare and to teach lessons, or parts of lessons. Learning by teaching should not be confused with presentations or lectures by students, as students not only convey a certain content, but also choose their own methods and didactic approaches in teaching classmates that subject. Neither should it be confused with tutoring, because the teacher has intensive control of, and gives support for, the learning process in learning by teaching as against other methods. 2016 Weng/Pfeiffer emphazises Martin as "a precursor of the frequently mentioned 'shift from teaching to learning'"
Students as teachers in order to spare teachers
Martin attempted to transfer the brain structure, especially the operating model from neural networks – to classroom interactions
Most teachers using the method do not apply it in all their classes or all the time.
Disadvantages: The introduction of the method requires a lot of time.
"Learning by Teaching" at Etica.AI
This concept, borrowed from education theory, seems to be useful even to force us to engineer systems in a way that is simple to pass on to colleagues.
Why
Potential method that helps to multiply teachers (interesting in initial environment with few humans of high experience)
Reinforces need for documentation (any type)
Reinforces need for simpler documentation
Many others
How
Use your human creativity to decide how.
See this playful example:
- I want to teach even a child how to manage stack of servers for I.A.
- Children do not like boring things. They prefer things with meaning.
- I try to use more images besides text commands. And video. Or other types of interactions.
- I try, along with technical explanations, to give motivations to use
- Focus on building blocks. Lego style.
A less playful example, very realistic as 2017 good practice, is using docker to manage complex infrastructure at the point that non-backend developers could run on they machines.
Agent-to-human
By agent-to-human for transfer of knowledge, means that the agent could be human-only, human-machine (human teaching to teach with machine assistance) or fully autonomous (yet a narrow AI) machine. But the target here, at least for this issue, means a human.
We do not have a full AI to use Learning by Teaching as efficient way to transfer knowledge compared to other methods (like machine learning) for machines, so agent-to-human (human-to-human, machine-to-human, human+machine-to-human) makes sense, but human-to-machine, at least for more than a very specific topic, not.
There are several ways to transfer knowledge to humans. Efficient ways of doing this are very important, even in a machine-intensive environment. "Learning by Teaching" is one way to do this that has advantages in the quality of learning at the cost of more time to do something.
Thinking about the future, it pays to prepare for technologies that will come instead of taking traditional methods of teaching in schools of our time. See also Zone of proximal development from Lev Vygotsky, that may explain why, even with vast knowledge on the internet, people do not learn without some kind of direction. Dale's Cone of Experience also have some, not fully, impact on retention level of knowledge.
Learning by teaching on Wikipedia:
"Learning by Teaching" at Etica.AI
This concept, borrowed from education theory, seems to be useful even to force us to engineer systems in a way that is simple to pass on to colleagues.
Why
How
Use your human creativity to decide how.
See this playful example:
A less playful example, very realistic as 2017 good practice, is using docker to manage complex infrastructure at the point that non-backend developers could run on they machines.
Agent-to-human
By agent-to-human for transfer of knowledge, means that the agent could be human-only, human-machine (human teaching to teach with machine assistance) or fully autonomous (yet a narrow AI) machine. But the target here, at least for this issue, means a human.
We do not have a full AI to use Learning by Teaching as efficient way to transfer knowledge compared to other methods (like machine learning) for machines, so agent-to-human (human-to-human, machine-to-human, human+machine-to-human) makes sense, but human-to-machine, at least for more than a very specific topic, not.