Organisational Mirror Neurons Gunter Heim, Aachen
|
||||||||||||||||||||||||||||||
| This site is an engineering fancy playing with some features of the connectionist paradigm of neural intelligence. I do not try to suggest how real brains might work nor do I try to suggest the best way to design knowledge work in a company. But I do feel like an engineer in the late eighteenth century might have felt when looking at cockwheels, steamengines for coal mines and a horse carriage. He might have felt that some of these parts might one day be put together to form some sort of steam carriage in which people could travel across the land at great speed and comfort, perhaps covering a hundred miles or more in a single day. The cockwheels, steamengines and horse carriages in this explorative sketch are neurons, synapses and human-machine-interfaces. | ||||||||||||||||||||||||||||||
Open Cast Mines as an Example To illustrate what I want to say, I use an example of knowledge work to be carried out in some open cast mines. Imagine a large hole dug into the ground to extract coal or ore. Such holes may be a few kilometers in diameter and a few hundred meters deep. In cases of heavy rain, all the water gravitates towards the deepest point, where the water level can rise a few meters in less than 20 minutes. This can pose a serious risk both to staff and equipment. Predicting the risk of such flooding involves a lot of different sources of information. The topographical features of the mine (which can change daily), wheather forecasts and the reliability and power of pumping installations may all play a role. |
||||||||||||||||||||||||||||||
Seven Types of Organisational Neurons There are seven kinds of organisational neurons that can be associated with knowledge work in goal-oriented organisations such as companies, ministries, universities or military units:
|
||||||||||||||||||||||||||||||
| The neurons described above are connected by data transmission lines. In moderate analogy to biological neurons these transmission lines are called axons. An axon can only pass on information from the sender neuron to a receiving neuron via a synapse. The state of a synapse determines the likelihood with which the transmitted data is actually taken into consideration by the receiving neuron. | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| The state of the
synapses can only be regulated by coordination neurons.
Their axons directly reach out to influence the synapses
of other neurons. Their activity can close or open
synapses:
|
||||||||||||||||||||||||||||||
| The figure above could be translated into (more or less) ordinary speech as follows: If the situation that there is a threat of the mine being flooded within the next 24 hours is true then the coordinator neuron will open the synapses that allow a transfer of data from the reporting list giving the location of the operating pumps in the mine to the working neuron. That may be a person or a piece of software responsible for updating the report and vice versa. If the threat of the mine getting flooded does not exist then the synapses between the report of the pumps and the person or piece of software that are symbolized on the right will be shut. | ||||||||||||||||||||||||||||||
| I now want to develop
some thoughts typical for the design of organisational
neural networks. Let it be assumed that a warning is
necessary if there is a danger of a flooding of the mine
within the next 24 hours. This is the first mirror
neuron:
|
||||||||||||||||||||||||||||||
| For the neuron to decide whether this threat exists or not, a number of statements are useful to check out. This gives a few more mirror neurons: | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| Looking a bit closer at the neuron in the middel, one may wish to know what indications of an impending pump failure may be. This leads to some more mirror neurons: | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| If this example was worked out in detail, one would soon arrive at a very large number of possible mirror neurons helping to assess the threat of the mine being flooded. These neurons can be interconnected with the help of axons: | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| I now want to add a few reports to show how the second type of organisational neuron may be used to model knowledge work within a company. If the top neuron in the figure above is on (i.e. true) it might be good to have an action plan at hand that gives some information on how to handle the situation. If the mine is flooded, it might be helpful, for example, to have the phone number of the mine manager at hand, to know whether the State Bureau of Mines has to be informed or not, or to know where to find out how many people are on shift in the mine and to check if everybody has got out in time. This information may be written in a report. Another report may contain information about the scheduled maintenance operations over the coming three days that might need to be postponed. And yet another report may give a quick overview of the location of all auxiliary pumps and spares. | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| The figure above shows how four of the seven types of organisational neurons can be used to model many aspects of concrete knowledge work that has to be done to identify and deal with the threat of a mine flood. The mirror neurons constitute the rationale that indicates the threat of flooding, the report neurons provide useful information needed in case of a real threat and the working neurons keep the report neurons` content updated. | ||||||||||||||||||||||||||||||
| The miner at the top of figure 4 is the fourth kind of organisational neuron, which is the motor neuron. It is one example of how the state of the neural network might reach out to into reality to affect the physical procedures and states in the mine. The miner might, for example, have a a small computer device with him that gives him an up-to-date overview of what is important to him. An open synapse of an axon coming from a certain neuron would mean that the miner`s attention is focussed on the information coming from that specific neuron. This can, for example, be realized by placing information from that neuron at the top of a list of all possible informations to be displayed. This is what the miner sees when looking at his portable computer device: | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| The screenshot above shows how the action plan for the case of a flooding of the mine is arranged at the top of the list of data at the left. The corresponding contents have a high likelihood of being transmitted from the source to the destination neuron (i. e. the miner). At the right of the screen one can see the detailed pieces of information that come under the heading of 'action plan'. This browser setting also marks the interface between the neural network and some motor neurons (i. e. human workers, machines, robots etc.). | ||||||||||||||||||||||||||||||
| The figure below introduces the loop for actually achieving learning processes. (Note that the neural company network is not closed in the sense that all relevant information sources are indicated.) | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
| Figure 6 shows how learning can be achieved in a company neural network. The coordination neuron continuously determines the setting of all the synapses relevant to the specific task of assessing the risik of an imminent mine flooding. The judging neuron produces quantified evaluations of how well these forecasts are working. If, for instance, there occurs a flooding of the mine without any prior warning the marks will be very low. Also, false warnings will produce low-marks whereas long-term warnings that appropriately fit an actual flooding will yield very high scores. These scores are logfiled against time and are being transferred to the learning neuron. The learning neuron also receives the settings of the synapses from the coordination neurons logfiled against time. Prosaically speaking, the dataset thus produced for a specific neuron may look something like the following: | ||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||
If interpreted, the table above may produce some tentative suggestions like the following:
Translated into the reality of the mine this might render a conclusion like: "For a good evaluation of the risk of a mine flooding, the considerations (synapse 1) of the shift leader of the water pump squad seem to be helpful whereas the advice of the foreman of the electricians (synapse 2) seems to be detrimental. The latter may be a show off who can well impress and bully people without substantially knowing anything about the matter. Information on the current windspeed (synapse 3) does not seem to be of much relevance either way. |
||||||||||||||||||||||||||||||
Conclusion I want to stop here with my intimation of the idea of mirror neurons and related neurons in the context of a neural sort of company intelligence. I think that the notion of synapses and neurons used to model some (out of many) aspects of organisational knoweldge work could help to implement semi-autonomous learning mechanisms that produce intelligence at a collective rather than individual level. The formalisation suggested above may serve to produce a more systematic unterstanding and active shaping of company knoweldge work, putting knowledge rather than business processes at the centre of considerations. Also, I think that once historic data of synaptic settings has been logfiled together with quantified evaluations of company success dreaming could become a heuristic strategy for large companies. Revelling through historic data and re-enacting past situations with the introduction of slight variations might open up a large search space for experimental new synaptic settings whilst being close enough to reality. During these phases of dreaming, company operations that require a reliable functioning of the complete neural network (like deciding on large contracts) may be shut down whereas some other functions that can be carried out more or less automatically to sustain the basic living functions (like carrying out running projects or handling legal affairs). |
||||||||||||||||||||||||||||||
| I believe that there may be great potentials in this way of thinking without the suggestions outlined above actually being useful in themselves. In the 18th century, many people played with the idea of turning steam into some mechanical use with the help of cockwheels, rods or pistons. Many of the ideas produced look ridiculous today. But out of this way of thinking was born the steam engine, coal-driven railways and much else that helped bring about the first industrial revolution. In this sense, trying to link up collective, company knowledge work with some aspects of neural information processing may also one day produce some sort of revolution - for good or for bad. | ||||||||||||||||||||||||||||||
Suggested Reading:
|
||||||||||||||||||||||||||||||
Some related sites in English: Constructed Brain: A test environment |
||||||||||||||||||||||||||||||
|
