Deterministic imprecision
Farmers in south India know that the monsoon arrives sometime in June-July. Crops have to be planted such that when the monsoon arrives they are neither too nascent, nor too old.
If for instance, there is heavy rain right after the seeds are sown, then they get washed away. On the other hand, if there is no rain for weeks after the seeds are sown, the crop becomes malnourished.
So, timing of the crop is critical.
Only problem is that, while the phenomenon of monsoon itself is deterministic, it is "imprecise". We don't know when exactly it will arrive and on what exact date will we get our first big rain.
This "deterministic imprecision" results in a number of collaborative and strategic activities among the farmers. They hedge crops so minimize risk, they cooperate with one another to reduce costs, and so on.
Deterministic imprecision is a characteristic property of nature. We can predict natural phenomenon at a coarse level. But we cannot predict specifics. In the colder regions of the world, we know that it snows in the winter, but we do not know when exactly and how much. We know for instance, a major earthquake is due in the Himalayas. But we don't know when.
There is perhaps a message in this deterministic imprecision. Deterministic imprecision is what motivates us to understand phenomena at a deeper level than at superficial levels. We need to build models of the weather. We need to understand risk. We need to understand costs. We need to understand needs. We need to prioritize. And so on.
In artificial systems, we seem to equate precision with quality. Specifications that are precise, are said to foster better quality work than specifications that are imprecise.
But usually what happens is that precision tends to foster "overfitting" to the specifics than towards meeting the spirit behind the activity.
Consider for instance, conference deadlines. Conferences put up deadlines in precise terms like 12 May, 23:59:59 PST. And what usually happens is that most of the submissions happen at the last moment. Web traffic peaks at this time and often results in disruptions and frazzled nerves.
The same thing is true with assignment submissions in classrooms. If the deadline for an assignment is set to precise terms like "Tuesday 1700hrs" then most of the students begin working on Monday night or Tuesday morning and submit the assignment very close to 1700hrs.
Such last minute work is primarily driven by a sense of compliance with rules, rather than adhering to the spirit of the activity (learning something through the assignment.)
Recently I've started to practice deterministic imprecision. I specify that the deadline is on (say) Tuesday without specifying the time. It is my prerogative to close the submission site on any time on Tuesday. If someone assumed that it was Tuesday 23:59:59 or something, well too bad. The rain has come and gone before you could till the soil..
If for instance, there is heavy rain right after the seeds are sown, then they get washed away. On the other hand, if there is no rain for weeks after the seeds are sown, the crop becomes malnourished.
So, timing of the crop is critical.
Only problem is that, while the phenomenon of monsoon itself is deterministic, it is "imprecise". We don't know when exactly it will arrive and on what exact date will we get our first big rain.
This "deterministic imprecision" results in a number of collaborative and strategic activities among the farmers. They hedge crops so minimize risk, they cooperate with one another to reduce costs, and so on.
Deterministic imprecision is a characteristic property of nature. We can predict natural phenomenon at a coarse level. But we cannot predict specifics. In the colder regions of the world, we know that it snows in the winter, but we do not know when exactly and how much. We know for instance, a major earthquake is due in the Himalayas. But we don't know when.
There is perhaps a message in this deterministic imprecision. Deterministic imprecision is what motivates us to understand phenomena at a deeper level than at superficial levels. We need to build models of the weather. We need to understand risk. We need to understand costs. We need to understand needs. We need to prioritize. And so on.
In artificial systems, we seem to equate precision with quality. Specifications that are precise, are said to foster better quality work than specifications that are imprecise.
But usually what happens is that precision tends to foster "overfitting" to the specifics than towards meeting the spirit behind the activity.
Consider for instance, conference deadlines. Conferences put up deadlines in precise terms like 12 May, 23:59:59 PST. And what usually happens is that most of the submissions happen at the last moment. Web traffic peaks at this time and often results in disruptions and frazzled nerves.
The same thing is true with assignment submissions in classrooms. If the deadline for an assignment is set to precise terms like "Tuesday 1700hrs" then most of the students begin working on Monday night or Tuesday morning and submit the assignment very close to 1700hrs.
Such last minute work is primarily driven by a sense of compliance with rules, rather than adhering to the spirit of the activity (learning something through the assignment.)
Recently I've started to practice deterministic imprecision. I specify that the deadline is on (say) Tuesday without specifying the time. It is my prerogative to close the submission site on any time on Tuesday. If someone assumed that it was Tuesday 23:59:59 or something, well too bad. The rain has come and gone before you could till the soil..
Comments