Emergence / Recognition

Emergence

Alan Levine writes, "I am looking for a metaphor or a real example of something maybe complex or just larger/valuable that does not arise solely from adding up many small components.

The meaning of a sentence does not arise solely from the meanings of the words that make it up. For example: “Look out” is a warning, but neither “Look” nor “out” contain any element of a warning.

A journey is not composed merely of many small trips. The very same set of small trips can either mean “a pilgrimage to Mecca” or not, depending on the intentions of the traveler.

A school of fish (or a murmuration of starlings) is not made up solely of individual fish; they have to be swimming as one unit, not merely happening to be in the same place at the same time.

A holiday is not merely a sequence of days with no work (the same thing could describe unemployment).

A song is not merely a collection of noises; they have to be arranged in a particular way, and they have to be pleasing to a listener.

A fact is not merely a collection of perceptions or observations (not even if joined together with logic).

Etc.

In general, reduction can be defeated by the following principles:

• function, where the whole has a function one of the original parts can have
• coherence, where the whole forms a shape or structure none of the individuals can form
• purpose, where the whole forms an intent that cannot be ascribed to an individual
• meaning, where the whole can have a sense or reference none of the individuals can have
• emotion, where the whole can evoke a response none of the individual parts can evoke

(This is not an exhaustive list)




Recognition

He also asks, Would you have any spare time in the not urgent future to share some thoughts as to how certifications might be done that are not badges or just by exam

I have thought about it, quite a bit.

All of the properties I allude to above (function, purpose, meaning, emotion, etc) are emergent properties. They depend in some way on the entities underlying them (this is known as ‘supervenience’) but they have properties none of the underlying entities have. This is exactly the case for knowledge, learning or achievement.

The mechanism for identifying emergent properties is recognition. For example, pattern recognition enables us to identify shapes out of a complex array of perceptions. Recognition is a result of the interaction between a perceiver and the entities being perceived.

Networks (specifically, neural networks, but arguably, networks generally) are recognition systems. We generally think of networks as pattern recognizers, but networks will recognize emergent properties generally. This is in fact how we evaluate achievement today (though we don’t always identify it as such):

- an aspiring doctor is observed by an expert doctor, who will assess the intern’s overall performance and recognize the intern as qualifying as a doctor

- a PhD candidate I given an oral exam by a committee, with the sole purpose of determining whether the committee members recognize the candidate as a peer

- an apprentice mechanic works under a supervisor, who will recognize the person’s expertise with cars

- the contests on Hell’s Kitchen present their plates to a committee who will recognize whether the food is of the highest quality

But there’s no reason why this needs to be restricted to individuals. A network can recognize expertise as well as an individual expert (and importantly, recognize different expertise). That’s the principle behind democracy (though of course we’ve seen it interfered with a lot).

It’s also the principle behind things like tagging, and behind the invisible hand of the marketplace, etc. It’s not perfect (human perception is not perfect either) and needs to be subject to constraints and conditions in order to be reliable.

I talk about it here: http://www.downes.ca/presentation/344