Unoutdone

In a recent post I wrote about the argumentative theory of reason, which proposes that people have evolved more to win arguments than to learn facts. At the end of the post I stated that although there seemed to be much cause for pessimism in the realisation that people would rather appear right and be wrong than to appear wrong and be right, I could see a way to address this basic cognitive shortcoming. I will attempt here to elaborate on my hope for how to make people learn in spite of their evolutionary imperative to avoid doing so when inconvenient.

The basis of my idea takes its inspiration from some Nietzsche that I first heard while an undergrad, which concerned the tension between competition and destruction. Nietzsche argued that struggle, including competitive struggle, made people stronger, but that people could compete not only by making themselves stronger but also by making others weaker. People could ‘undo’ each other as well as ‘outdo’ each other. Nietzsche had no clear answers as to how outdoing could be made to supplant undoing, and I do not know who else has come up with any good answers.

It strikes me that one way in which a person undoes another is by solving a problem for that other that said other might have solved for themselves had a solution not been given to them. Solving problems for people may not seem like a way to harm them, but harm can be done to a person by deflecting them from opportunities for growth, to the extent that the deflections culminate in the helped person concluding that they could not have helped themself and cannot in general help themself. This has implications for relationships between teachers and learners, and at the same time extends the idea of the teacher-learner relationship beyond formal education to social transactions in general. Essentially, someone that provides you a service also teaches you a lesson- that the way to meet the need that the service provides is to rely on the service’s provider. The inseparability of the service and the lesson can be understood in terms of connectivist learning theory, which models the possession of knowledge as the ability to locate and apply the capabilities of networks to the achievement of goals. Experience of purchasing services necessarily incorporates the navigation of socioeconomic networks for goal related purposes.

In humanity’s evolutionary past, people would have tended to be much more stark than we generally are today in appraising the extent to which the provider of a service could be relied on to provide it again in the event of some future need for it. Materially advanced cultures have heavily incorporated the principle that services will in general be available as and when required for those who can pay for them, emphasising the primary importance of being able to pay for services over any considerations of how difficult or otherwise it might be to find alternatives to purchasing the services. Under conditions of a default expectation of inability to cope with the need to find alternatives to services whose provision has become taken for granted, service providers are incentivised to artificially increase the obscurity of the methods and techniques involved in provision of their services. A seller of services does not want a buyer of those services to know how the services actually get provided.

Limits exist to how mysterious it is beneficial for service sellers to make their methods. Consumers can be distrustful of what they know and understand nothing whatsoever about, so a successful service seller aims to present consumers with pseudo-explanations that reassure the consumer that the seller understands the methods, but that those methods would be too troublesome for the consumer to attempt to understand. It saddens me to say that formal education programmes can exhibit apparently similar design principles to those of service providers in general. Educators have plenty of pseudo-explanations to offer that seem to exist at least partly to reassure learners that their teachers will have answers for them later that they cannot provide on demand, and that this unavailability is down to failure of learners’ understanding, correctable of course by the purchase of additional educational services.

But how then is the spiralling growth of undoing that technocracy and capitalism incorporates to find room for the growth of useful outdoing? My contention is that a sort of re-weighting of incentives is called for that makes people think it more worth their while to learn how they can solve (and even to define) their own problems than it is to engage in exchanges with those offering solutions. My further contention is that the opportunity for such incentive readjustment is rapidly approaching in the form of large-scale technology-driven unemployment, as exactly the predicament of large numbers of people being in the position of no longer having anything to exchange for services once their labour loses its value is the kind of drastic incentive for people to look for alternatives to purchasing solutions that has the potential to make people start valuing their own efforts to solve their problems.

The sweeping (although possibly quite gradual) changes of the kind I envisage are conditional largely on three points. The first condition is that groups of people have a capacity for a crisis-induced resilience that is not necessarily apparent from their behaviour under circumstances that appear to guarantee their continuity. The second condition is that widespread technological unemployment is actually imminent, and the third factor is that viable alternative methods of solving peoples’ material problems can be implemented in the timescale associated with the onset of pervasive technological unemployment.

There have been various examples throughout history of unexpected collective organisation in the face of adversity where surprising levels of resilience have been demonstrated. In the recent history of the UK the obvious examples come from the Second World War- the so called spirit of the blitz and the Dunkirk spirit. The first example exemplifying a community taking action to preserve itself in spite of inadequate or counterproductive official policy (The authorities in London during the blitz initially prohibited the use of underground train stations as shelters. People  simply ignored this restriction and made the stations foci of communal support and resilience), and the second example exemplifying the distributed coordination of large numbers of agents acting effectively to cooperate on a shared goal despite facing dangers and challenges that they were not trained for using tools for purposes that they were not designed for. More recently, When the city of New Orleans was struck by hurricane Katrina, the inadequate and delayed government relief efforts stimulated various cases of spontaneous self-organised adaptive responses from many ordinary people who recognised that they needed to take action to save themselves and each other and saw that the assumption of conventional social continuity was no longer something to be relied upon. The history of the modern state of Israel could be considered as a case in point; a country originating in crisis that has never actually concluded- just receded and advanced unpredictably- has made itself into what has become known as the startup nation. To what extent it is a general feature of human society to exhibit this kind of hormeosis is of course debatable, but it is clearly not out of the question. 

It is also a matter of considerable debate just how much technological unemployment to expect and over what timescale. Estimates vary greatly, with some commentators predicting an overall growth of job opportunities and others predicting billions of today’s children facing a lifetime of zero or greatly reduced earning potential. The possibility of low paid, de-skilling of work is probably more immediate than actual unemployment. Plentiful work may continue to, exist but what that work will pay may depend ultimately on the cost of automating it, regardless of what workers might need to be paid. Falling pay could presumably be compensated for by falling production costs associated with a low cost workforce, but how well these two factors might balance each other out is anyone’s guess. One way or another, a large part of the current global workforce may come to be in a situation where the exchange value of its labour is insufficient for sustenance.

What though, if rather than acquire sustenance (and hopefully more) by exchanging individual labour for what is produced by collective labour (however much supplemented by automation), people did more to produce for themselves what they required? For this to be possible, methods of production would need to be drastically scaled down and simplified in terms of cost and usability and would need to be made much more adaptable and flexible. Localised, bespoke production methods are currently rapidly developing. The most well known example of this is 3-d printing (3-d printed houses made in one day for example), but similar new maker technologies involving biochemical production have perhaps even more transformative potential, as do open-source, cloud-based machine learning tools for developing control systems. All the parts exist or are rapidly approaching existence that suitably combined make possible the production and maintenance of the kinds of goods that until fairly recently would have involved an elaborate industrial infrastructure and concomitant technocratic governance and regulation far removed from the comprehension of the consumers of the goods.

While production technology has remained a black box for consumers and the products generated by these technologies have been obtainable in exchange for labour, consumers in general have had no significant incentive to attempt to understand production processes (or distribution processes, or logistics in general) and the technological and scientific knowledge upon which they are based. The threat of unemployment and obsolescence stemming from rapid technological change has unsettled the equilibrium of industrial consumer society. The adjustments required in terms of retraining and re-educating workforces is one that such workforces are unprepared for and have no obvious path toward adaptation to.

How can such a sudden shift in learning culture for so many be achieved in the timescale  required if it is the case that people’s evolved predisposition is to be motivated to win arguments rather than to learn factual knowledge? People will not readily endeavour to learn what they do not believe that they need to know. Trying to persuade people that they do need to know something that they assume they do not need involves arguing with them that this is the case and winning that argument, and the easiest way for people not to lose that argument is not to engage in it in the first place.

I have come to think that the way out of this bind is to give people easily relatable incentives to engage in arguments that are concerned with production techniques, namely by giving them opportunities to produce things that they would have otherwise had to work to earn the money to buy. Some sort of carrot and stick model of subsidies and tariffs, which could only really be governmentally enforced, could be applied to goods and services such that incomplete forms of them that required some finishing, customisation, and tuning by the consumer would be made much less expensive than fully finished versions that could be used immediately without acquiring any knowledge of how they worked. Basically, a person could acquire consumer items that they would otherwise not be able to afford if they were willing to engage with the manufacturing process of those items, ideally as part of a community of producer-consumers (I think that the word ‘prosumer’ has already been devised) configured to facilitate arguments over optimal production methods. The more deeply a consumer was involved with the manufacture, and by extension with the design, of consumer items, the cheaper they would be able to buy them. The logical extension of such an incentive structure would be prosumers initiating new consumer items and developing them from existing items or even from raw materials, gradually moving away from sharing partial knowledge of production methods to buy things cheaper to gaining comprehensive knowledge of production methods in order to make things without having to buy them at all.

There are a lot of ifs and unknowns in the scenario that I put forward here, but it is a scenario that offers a hopeful response to fears of human obsolescence and runaway technocracy. 

Connectivism as an Automation Ethic

The purposes of education are dependent upon the character of the society in which education occurs. In human prehistory, education must (presumably) have primarily existed to preserve and transmit practices that had survival value. Such practices plausibly enough included rituals and traditions that helped reinforce the tendencies to cooperate within groups, as well as directly utilitarian skills and knowledge sets. Education was practiced to produce detailed knowledge of specific local ecologies. Social value was awarded to those who were able to adaptively apply such knowledge for the benefit of their communities. Societies valued individuals with well attuned heuristic and implicit knowledge of their environments- wisdom for want of a better word.   

In the historical era during which mass compulsory formal education started to be introduced, the purposes of education had become highly specialised for supporting the growth of industrial workforces. Education was practiced to produce workers who could behave predictably, reliably, and interchangeably. Social value was awarded to those who were able to bring about material productivity. Societies valued those who worked hard and who were relentless in discovering and applying ways of increasing the efficiency with which productive work could be done in order to maximise production. Contemporary educational institutional structures and procedures are to a considerable extent legacy systems of the mass education model created for the age of industrial growth. Mass education was designed to produce assembly line workers and it prioritised developing individuals’ competences at dependably repeating standardised procedures without variation.

Learning behaviour in humans is in general much more closely aligned to the prehistoric era than the industrial era. Human learning behaviour is in general an experiential set of implicit heuristics- informal, nonlinear, undirected, and often unreflective. The iterative fine-tuning of a person’s set of implicit heuristics for navigating the complexities of existence (their ‘wisdom’) constitutes the great bulk of what people learn throughout their lives. When formal educators refer to learning, they do not generally refer to this sea of invisible learning, but rather to an archipelago within that sea of relatively stably-defined island-disciplines. Formal education can be thought of as the flow of traffic between these islands, but learning in a more basic and pervasive sense is the total motion of the wider waters. Everyone is learning throughout their lives, unavoidably, although they may not realise it, and it is this ubiquitous lifelong learning process that is one side of the barrier to engagement with and success in formal education as it currently exists.

Why should it be that a human instinct to learn should tend to be an obstacle to formal education? To understand the roots of this schism it is useful to consider what would plausibly be expected to characterise human instinctual learning. The first and perhaps most obvious point is that human learning is adaptive and flexible. Humans do not automatically repeat inherited behaviours without undergoing a process of behaviour acquisition, and that process is heavily shaped by circumstance and contingency. Much of human behaviour is the result of complex interactions that can have unpredictable outcomes, and these behaviours can be of critical survival value. For humanity to have survived successfully, some mechanism must have guided the complex interactions steering behaviour acquisition that selects for survival promoting behaviours and against behaviours detrimental to survival. Put plainly, because humans make decisions by experimentation rather than rigidly preset automatic responses, something must make humans recognise and prefer successful experimental outcomes- people must be intrinsically motivated to correctly understand their world otherwise they would not long remain in it.

It is an encouraging premise that people intrinsically want to learn, but considered on its own it has the profound limitation that it only reveals what an individual’s motivation to learn is in a world where so much of knowledge is cumulative and depends on shared processes. The process of collective learning, and by implication, of teaching, brings to bear far more double-edged possibilities than the basic human will to understand their world. Learning as an activity associated with communication between humans is inevitably shaped by more general principles affecting human communication and relations. When one person communicates something to another, the possibility exists of the communication being a vehicle of persuasion. Just as harsh evolutionary logic dictates that to survive humans must have developed an interest in successfully understanding features of their world, the same logic even more irrevocably determines that the capacity for a person to use their communications with others to persuade those others to do what best serves the person’s survival interests, and conversely to be less susceptible themselves to other’s manipulations, is a capacity that should strongly selected for. When an offer is made by one person to teach another person something, it is therefore an understandable response of the erstwhile student to wonder what motivates the teacher to make such an offer, and to be appropriately suspicious. The intrinsic satisfaction that comes from a fuller understanding of some aspect of the world is inevitably coloured by a student’s ruminations on how much that understanding may serve to manipulate their perception of that aspect of the world.  

On the basis that learning from teaching, as opposed to solitary self-guided learning, is predicated on the existence of at least some threshold level of trust in the teacher by the student, it is natural enough to wonder what factors are likely to help in engendering trust in teachers and what factors tend to undermine such trust.

The most self-evident factors conducive to forming and to eroding student trust in teachers relate to students’ perceived viability of the methods and techniques used by teachers, the integration of these methods with the desired learning goals of the students, students’ perceptions of the authenticity of teachers’ expertise, and the extent to which it is students that initiate and terminate the learning process.

When it is stated that there exists a human impulse to understand the world, the understanding referred to is of the type that is much more heuristic than analytical. The understanding that integrates easily with the basic learning impulse is woven from imminent considerations and from assumptions that such considerations are associated with. This kind of understanding is not sceptical but rather tends to be confirmation seeking, searching for connections between existing assumed knowledge rather than positing apparently unconnected alternatives, although incorporating open-ended experimentation when lacking clear direction. This preference for constructing new knowledge through linking existing knowledge extends to a preference for linkage between means and ends. Learning activities which are explicitly undertaken to address specific desires and which recognisably resemble the actions that would be taken to fulfil such desires more easily meld with the basic human learning impulse.

The predilection for symmetry between means and ends applies not only to individual learning but also to learning from teachers. Teachers able to demonstrate that they have themselves mastered what they are teaching have less need to persuade students to accept their guidance, and so do less to raise students’ resistance to persuasion that interferes with the flow of their learning instincts. Less resistance is provoked also if the teacher’s motivations for teaching are easily interpreted by the student as a transaction between teacher and student that is based on mutual consent.  

If the norms of contemporary education systems are analysed in terms of the considerations of factors affecting students’ trust in their teachers then the shortcomings of most education systems relating to all these factors is very clear.

Firstly, institutional education heavily leans towards analytical methods and techniques, to the extent that to be formally educated is almost synonymous with being trained to abandon heuristic methods in favour of analytical methods. Effective analytical thinking skills tend not to be very successfully inculcated by institutional education, with various static forms of analytical thinking instead being learned as given truths that are ostensibly taken as correct descriptions of the world but which only very superficially replace students’ learning heuristics, which continue to develop in the background, confusing both the learning of analytical methods and heuristic methods.

Secondly, the curricula of institutional education are very strongly aligned with analytical approaches to distant, long term learning goals. These curricula are highly modular and hierarchical and only after many cumulative steps do they recognisably approach the mastery of a skill the possession of which is expected to be a condition of self-sufficiency in a post-educational existence. Students in institutional education therefore not only struggle to understand how what they are doing will result in their useful learning but also struggle to understand what the use is of what they are struggling to learn, and ultimately why they ought to learn it.

Thirdly, the existence of teaching as a profession means that students do not in general learn from people that exemplify the condition of self-sufficiency in a post-education system existence. Teachers are (for the most part) not instructing their students in how to become teachers. Education professionals are all too easily conceivable by students as appendages of an education system, as having never left that system, even as lacking some ability or desire to leave it. Professional teachers can easily seem to students to be teaching them only how to be students and not to be the self-sufficient individuals that they will be required to be after their education. It is understandable that students may doubt that their teachers genuinely know how to effectively prepare the students for the world awaiting them after their formal education ends.

Fourthly, the compulsory aspect of institutional education greatly exacerbates the resistance to trust in the education process that the other factors produce. The lack of a student’s choice in whether to partake in institutional education can have the effect of students coming to think of their teachers as monopoly-owning rentiers, mere middlemen that have assumed the role of gatekeepers to the access of opportunities to participate in activities from which the understanding that they desire to obtain is assumed to emanate. The increasing correlation of study to the acquisition by students of debt surely reinforces this perception.     

For much of the (fairly short) history of institutional education applied at the mass societal scale, these shortcomings education systems were tolerable because the systems did not ultimately require the establishment of students’ trust so much as their compliance in performing according to simple systematic instructions, and the ability to obtain that compliance was supported by various other cultural institutions. In more recent times however, the character of society, particularly in economic terms, has undergone a set of transitions that has resulted in an unsustainable degree of incompatibility between the type of learning that educational systems are able to provide and the type of learning required in the emerging socioeconomic paradigm.

In the current era there has been a shift in emphasis in the interpretation of efficiency of work. Efficiency understood as the leveraging of division of labour into standardised processes in the pursuit of ever greater capacity for productive work is being superseded by efficiency meaning the ability to do only the optimal amount of work in the achievement some particular goal- efficiency as not doing more than is actually necessary or less than is required to achieve a goal; working smart rather than working hard. Societies in the world of the near future will conceivably value individuals who combine a kind of intuition resembling the wisdom of prehistoric people with a propensity for the discovery of ways to increase efficiency. I suspect though that the resolve to work hard will not endure as a greatly valued trait. A person will not be valued for their ability to do hard work but rather for their ability to find ways to automate hard work. Hard work will come to be seen as the price of failure to automate, as much as scarcity has been seen as the price of failure to work hard, or longer ago, seen as the failure to be wise.

In a somewhat similar way to the way in which efficiency has started to be reinterpreted as having more to do with parsimony of effort than with amplification of production, the meaning of automation may start to be interpreted in a new way. The wisdom required for success in the prehistoric world, and of pre-industrial worlds generally, can be understood as the ability to recognise and appropriately respond to ecological processes occurring independently of human control. Processes that are not human controlled are in human terms automatic, in the sense that no person needs to (or in this case, no person can) induce their occurrence any more than a person can prevent them from happening. Fruit growing on wild trees is in this sense produced automatically. If people have to pick fruit, then the fruit picking is non-automated work, but if the fruit can be made to drop by a slight shaking when they are just ready to fall and are caught in nets, much of the work of picking has effectively been automated, especially if the shaking is done by the wind rather than by human effort. Knowing when to spread nets and under which trees is effectively knowledge of how to automate a large part of the fruit gathering process. Automation of processes has tended to imply not only that people do not themselves fully perform the processes, but that the performing of the processes can be initiated and terminated arbitrarily according to a person’s volition. The requirement for processes to be made independent of natural timescales and orders need not be an important one if the processes involved are sufficiently diverse and numerous that required levels of production can be met from the available set of naturally ordered processes occurring at any given time.

Processes that consist of human activity can be automated in the sense of automation as taking advantage of some occurrences that did not require significant specific inducements to bring about. This generalised interpretation of automation refers to any sort of approach to achieving goals that relies mainly on exploiting various sorts of inertia; using the existing dynamics of some system to act in a certain way to achieve a goal rather than expending effort on altering a system’s dynamics.          

In some ways, the distinction between valuing hard work and valuing making the results of hard work easier to achieve could also be applied to the work ethic of developing industrial societies. The productivity that a person could bring about was ultimately the most fundamental way that their value to an industrial society could be measured, more important than a person’s basic willingness to work. The attitude that a well thought of person ought to be personally industrious and to value work for its own sake continued to be seen as important however. The perceived intrinsic value of hard work has for some time now though been eroded by the shifting understanding of efficiency. An expression of the redefinition of efficiency, and of automation, can be seen in the economic tendency for seeking increases in production to be substituted with seeking increases in services and also increases in accumulative processes.

The basic distinction between an accumulative process and a productive process is that an accumulative process is transference of some pre-existent commodity between parties and such an activity does not necessarily require any activity to occur that adds to the sum of that commodity. Manufacturing represents a basic productive activity and a comparably basic accumulative activity is trading. When people trade, they are not directly taking action to increase the sum total of whatever it is that they are trading in, although they may very well be incentivising that action to be taken. Accumulative activity can take notably predatory and exploitative forms. An egregious example is the perpetuated collection of debt interest payments that greatly exceed the principal borrowed and which the requirement to regularly repay prevents the borrower from ever saving sufficiently to be able to pay off the original debt. The establishment of monopolies that create barriers to entry into markets are not themselves accumulative activities but they limit the scope of who is and is not able to act as an accumulator, and so understandably monopolistic practices tend to be correlated with accumulative practices. These kinds of expression of accumulation are so nakedly parasitical that it is not difficult to recognise them as ultimately unsustainable. They act as disincentives to production, much as production was disincentivised prior to the industrial era by accumulative state powers. Why would someone make the effort to increase production if any surplus is accumulated by someone else?

A commodity that is currently being rampantly incorporated into accumulative processes is data about a multitude of systems, including systems of human organisation. The collection of very large datasets and the analysis of this data using machine learning methods is the engine of the emerging interpretation of efficiency as the knowledge of how to interact with systems so as extract the greatest benefit from the interaction at the lowest cost to the accumulator. Using big data accumulatively is tantamount to forecasting the future and placing bets on what will happen, where these bets are expressed as marketplace transaction offers. These transactions originally tended to be in ownership of commodities and derivatives of these commodities and were carried out by professional traders, but increasingly transactions are in rental of commodities and in services and are carried out by general workers as an aspect of their work. The entirety of the so-called gig economy and sharing economy can be modelled as an ecosystem of agents offering goods and services to each other where each agent attempts to minimise the work that they are doing and simultaneously maximise the work that the rest of the ecosystem can do for them, automating their own work and accumulating the work done by others (albeit temporarily). People that own goods that they do not use for significant intervals can profit by charging other people for temporary use of those goods, and if such transactions occur then the demand for the production of such goods is reduced. Someone who makes a commute in a vehicle with unused capacity can profit by taxiing customers that want to travel along some parts of that commute, thereby reducing the demand for more vehicles.

Accumulation is not in itself productive, but can incentivise production if producers retain a portion of what they produce that is sufficient to incentivise them. Accumulation by big data and machine learning is not in itself productive, but it has the potential to vastly reduce the amount of production required by human societies.

In data and machine learning driven accumulation, an accumulator uses the work done by producers to allow the accumulator to require less production, as the accumulator reduces their expenditure requirements by reducing the inefficiency of their expenditure; learning how to make less costly choices. The accumulator benefits from knowledge gained about producers, without making a contribution to the production. This accumulation does not directly apply any costs to producers though, and if the beneficial knowledge accrued by the accumulator is in some way shared with producers, even indirectly, then producers can also gain by reducing their expenditures. The great exemplar of this process is the company Amazon. Some production is still required however efficiently existing resources are utilised, so accumulation cannot become the entire basis of economic activity, but it may become plausible to reduce the baseline level of production required to a level that can be met almost entirely by automated processes.

Productive work of any kind is, at least in principle, to some extent automatable. The extent to which some productive work cannot in practice be automated can be considered as a lack of knowledge as to how to automate that work. The lack of knowledge of how to automate some work can be thought of as a positive quantity, a quantity of ignorance as to how to fully automate the work. If ignorance of how to achieve full automation is considered as a positive quantity, then it can be seen that the reduction of such ignorance is equivalent to the consequent automation of the work- at least potentially. What this implies is that data and machine learning driven accumulation is a mechanism for expanding the potential for ever greater automation of productive work, and hence that a sufficiently great degree of sufficiently distributed accumulative processes should result in the amount of productive work that ultimately has to be done without automation, and therefore by people, becoming ever smaller. The final limit to how much productive work can be automated by accumulative processes is determined partially by the combined knowledge of the systems contributing to the process. To effect a significant reduction in the amount of productive work that remains for people to do, a vast body of knowledge related to productivity efficiency improvements must already exist, as it in fact currently does.

Machine learning, as its name suggests, is automated learning, in the sense of learning as the ability to bring about desired goals. Applications of machine learning to large sets of data have over a relatively short timescale become enormously important to the effective automation of networks for accumulative economic processes. To a lesser extent machine learning has also been successfully applied to the automation of increasing the understanding of the operation of natural systems. The techniques of using machine learning to process datasets are, for now at least, relatively freely accessible and not irrevocably bound up in monopolies. Machine learning algorithms can constitute intellectual property, but there appear to be significant limitations on how difficult it is to prevent an existing algorithm’s behaviour from being modelled by another algorithm. As machine learning algorithms are not designed to achieve specific tasks but rather to learn how to achieve specific tasks, such algorithms are by nature hard to pigeonhole as being a machine for doing x, as intellectual property laws would find convenient to deal with. The fact that how machine learning algorithms produce the behaviour that they do in fact produce may not in practice be determinable in any straightforward way makes the judgement that a particular algorithm is or is not a copy of some other algorithm a problematic one. The ownership of data is perhaps a field where accumulative practices could more easily become excessively predatory and exploitative, but as with machine learning algorithms, this has not yet happened and may not do so. For as long as data and the machine learning processes continue to avoid being made artificially scarce, accumulation using data and machine learning, both from the natural and the human world, have the potential to become dominant socioeconomic driving forces.

To what extent do current education systems prepare learners to devise ways of applying accumulative processes to automate and minimise the need to do work? I suggest that the answer is that there is little to no formal provision by education systems for such preparations and comparably little scope for the movement toward adoption of approaches conducive to such preparations. I do not consider that this failure is due to an intrinsic difficulty in reconciling educational processes with the acquisition of an automation ethic, and in fact educational processes designed to be in alignment with automation ethic acquisition have the scope to be extremely well aligned indeed, more so indeed than the industrial era education systems have been for aligning learners to the work that they have been prepared for. Connectivist learning theories are the basis of the potential for a superior alignment.

Connectivist models of learning consider learning to consist of the ability to navigate networks within which existing knowledge is distributed, and to form new connections within such networks. Connectivist learning is social and collaborative and is defined in operational and functional terms. Knowledge is understood in terms of the ability to produce some specific outcome and a learner is considered to possess knowledge to the extent that they can draw on networks to produce desired outcomes. Crudely, knowing something is equivalent to knowing where that knowledge resides and how to access it. Where knowledge is purely understood in terms of the capability of attaining goals it is meaningful to think of networks as being machines for automation of the process of attaining such goals. Connectivist theory incorporates as a foundational premise that knowledge resides in nonhuman as well as human entities, and this knowledge need not merely consist of static data but also of dynamic processes that act on data; as is the case in machine learning systems. Connectivist learning processes clearly have strong parallels with machine learning based automated accumulative processes. The connectivist paradigm’s parallels with the economic paradigm arising from machine learning based automated accumulative process implies that connectivism can not only determine how learning occurs but also the bulk of the content being learned, producing a powerful synergistic effect that existing educational models are in comparison very limited in their ability to produce. The problem in contemporary education of learning practices having only weak relevance to the activities that education ostensibly prepares learners for need not apply to sets of learning practices that are in many ways clear precursors of and simplified versions of the activities that education prepares learners for.

A practical appraisal of what activities learners would plausibly be explicitly motivated to engage their attention on suggests that such activities would need to be related to problems of enough relevance to learners that they wished to solve them, be problems which learners did not already fully know how to solve, be problems which learners could conceive of themselves as being able to solve, and be problems which learners did not expect others to fully solve on their behalf. Thinking in these terms, what kinds of educational practices based around connectivism could plausibly develop? I speculate that two main types of practice would be expected. These can be summarised as consisting of distributed machine learning development projects and of data generation virtual environments.

Distributed Machine Learning Development Projects:

Groups of learners could collaborate on learning projects that aimed to solve specific problems, which is to say to automate the achievement of specific tasks, through the use of machine learning algorithms working on datasets. The types of problems addressed could easily be made quite open-ended and driven by the interests of learners. The projects would involve distinct but overlapping sub-tasks related to data collection, data evaluation, the selection of appropriate algorithms, training the algorithms with data, evaluating the success of the training, and evaluating the economic potential of the systems developed, as well as working with robotic systems for automating physical rather than cognitive work. Learners of diverse ages, abilities, and skills could contribute to such sub-tasks. Even very inexperienced learners could be usefully involved in collection and of comparison of data. Much of the learning involved in these kinds of projects is closer to the ‘wisdom’-like informal learning of primary familiarity to many learners, while ample scope would also exist for more analytical treatments of the principles of machine learning and of the systems being modelled by machine learning systems for those learners inclined to engage with such considerations. Projects’ outputs could result in the generation of authentically useful problem solving systems and in the collection of useful data, and even if such utility was low then learners would be directly acquiring familiarity with how to participate collaboratively in projects aimed at the development of similar systems that were more likely to be useful. The relevance of such learning activities to economically useful activities is obvious, and it is easy to imagine employers and entrepreneurs seeing benefits in becoming involved with projects undertaken for educational purposes and be willing to act as teachers, providing strong exemplars of teachers as people who also do what they teach.

Data Generation Virtual Environments:

Before projects about problem solving can be undertaken, learners need to have ideas concerning what problems to try and solve, meaning they must have some awareness of various phenomena and knowledge about these phenomena. That which has already been learned can be considered as consisting of sets of data. Presenting existing data to learners in ways that are conducive to engaging their curiosity can be assisted by giving learners flexibility over what to explicitly learn about while simultaneously exposing them to information about a vast and diverse range of phenomena. Emphasising the connections between such phenomena can be a way of maximising the chances of learners encountering the phenomena that they will have the strongest propensity to want to learn about. The greater the diversity of options, the less prescribed the learning would be and so the less persuasion required to continue it and the less resistance from learners to learning. Furthermore, where the phenomena available to be studied are massively diverse, more opportunity exists to discern points of comparison between them and for this to lead to the recognition by learners of transferability of their understanding of general principles to different phenomena, including those related to generic machine learning system principles. Because there is at least some minimal residual learning impulse in even the most apathetic or resentful learner, if a learner is presented with a fluctuating stream of stimuli which they have the option to influence their exposure to, then given sufficient exposure to the stream of stimuli, a learner will inevitably start to respond with some sort of preferences for some stimuli more than others. If the sum of collected knowledge were represented in some sort of virtual world and learners entering that world were propelled throughout it in a way that was basically random, sooner or later learners would develop at least implicit preferences for some parts of the world and exercise choice in moving toward preferred parts and away from unfavoured parts. The choices made by learners in this kind of virtual world would generate potentially useful data concerning what learners are curious about, which can indirectly provide information about what the learners understand. Importantly, the virtual world of knowledge would contain multitudes of hyperconnections that would greatly reduce the chances of learners staying in only certain limited regions of the world and not finding how its parts relate to each other. The emergent navigation decisions made by learners would be better described as exploring than as learning, but they would represent an at least implicit learning of the connectivity of the parts of the knowledge space. Exploring knowledge spaces can be thought of as engaging learners’ default mode cognition, also referred to as ‘mind wandering’, to promote generation of novel and creative associations, while at the same time immersing learners in information, some of which they would inevitably develop familiarity with.

These kinds of connectivist learning processes need not supplant other kinds of learning (although for some learners that might perhaps be appropriate) so much as supplement them. It would ultimately become an economic question as to which learning activities were considered more valuable, decided by the employability (including self-employability) of learners educated along different lines.