In the work, Kuhn does the intellectual spadework for the sociological analysis of science – the idea, contra Karl Popper, that scientists are deeply concerned with their own existence inside a community.  Kuhn argues that mature science (typically this means the physical sciences) is defined by the hegemonic influence of a single instrumental perspective – what Kuhn dubbed a paradigm (thus plaguing us all to the misused cliche, “paradigm shift)

Under paradigms, scientists are given a shorthand for determining how to conduct their discipline – what questions to ask, what apparatus to use to answer it, what data to ignore/highlight etc. Paradigms frame the discourse within the scientific community, giving a common referent for what are often highly abstract concepts.

Kuhn argues that paradigm adoption is rarely a rational, evidence based process of considered migration. It is instead, usually the result of a crises in the old paradigm – scientists clamour aboard the most plausible alternative in the hopes to be saved from the problems plaguing the old paradigms.

In the word of Herbert Simon, they “satisfice” in their selection.

It takes little effort to see that Kuhn’s theory hinges upon many different assumptions about human cognition. He doesn’t make a normative argument about what science should be, nor does he make categorical claims about what science can never be – he merely states that paradigm adoption/rejection is how science operates currently.  That is, how science is now supervenes on truths regarding how human beings think about science, which in turns supervenes on how human being think at all.

This has some interesting ramifications for cognitive science and organizational cognition. Science is arguably history’s greatest example of collective cognition and many scientific enterprises are exemplars of successful organizational cognition. Kuhn’s description implies strongly that human cognition strictly determines collective human enterprise.

If our best example of individual human intellectual transcendence (science knows more about the world than any other collective enterprise and has the clearest grasp of what it knows) is confined quite dramatically by certain very human, very mundane cognitive realities what implications does this raise about the possibility space for future collective human efforts?

It does make a strong case for something I’ve believed. That organizational systems are surprisingly regular at very large scales and the relation between individual and collective is less complex than the system’s scale would indicate.

This issue is akin to certain problems in physics. A system containing one or two particles is easy to make predictions. Another system containing millions of particles is also, at the systemic level, easy to predict. It is in the intervening levels of complexity that prediction becomes difficult.

This is common-sensical when one considers the difference between making predictions regarding one’s graduating class vs. one’s sibling vs. one’s generation.

The critical difference here is that it is clear what kind of predictions are possible at the uppermost and lowermost levels of complexity – the kinds of things one can plausibly say of a generation or a person respectively.

Kuhn’s insight is that the uppermost level has a character that is, surprisingly to many, heavily constrained by the lowest level.

This view of paradigms, in many ways, parallels attentional scaling. That the individual elements one retains in working memory  strongly determine the range of possible gestalts one can perceive. A new gestalt is almost always the result not of new insight on static elements, but of the shifting nature of the elements themselves – by foregrounding one fact.

To Kuhn, paradigms are distinct from gestalts in that paradigms are one-way. Gestalts, according to Kuhn, can be manipulated and flexed at will.

This seems convincing when one considers the example of Necker Cubes or Navon letters. However, in the light of insight problem solving, this distinction becomes problematic.

In issues of insight problem solving, the solution once discovered becomes permanently salient. To the extent one can recall the solution to say, the nine-dot problem, one can’t help but perceive the entailed salience map.

Thus, paradigms can be seen as the salience maps generated from the solution to certain ill-defined problems that plagued the previous paradigm. By making this adjustment to Kuhn, one can square his theory with many important psychological phenomena.

David mentions that critical to a politician’s success is the filtration capacity fulfilled by his aides and advisors – without people to intercept and cognize the fire-hose of information, picking out what is relevant,  politicians would be paralyzed.

This issue of relevance realization is the central problem of artificial intelligence, and is fundamental issue for all of cognitive science.

Of note, this problem is arguably the fatal flaw in US DoD’s mostly defunct NetWar theories espoused by Nagl and other “fourth generation warfare” (4GW) theorists, especially those affiliated with the air-force and navy. Military decision makers, now flooded with unprecedented quantities of information from the sensor-net swarms, could not effectively coordinate action in the face overwhelming equivocal facts.

This need for a relevance realization capacity amongst politicans can help explain the sloganistic over-simplification of political opinions.  Policy solutions need to be pigeon-holed to be comprehensible under the cognitive load typically borne by decision makers.

Access to the morass of opinion that is the web, as David observes, could have utterly devastating effects over the fragile, centralized command-and-control structures seen in contemporary political systems.

Thanks to modern technology, leaders have access to extremely detailed information about practically any aspect relevant to their responsibilities. Consequently they often micro-manage many aspects of the decision making process by the reinforcing feedback of information monopoly and performance superiority. They make better decisions faster than their reports because they have access not only to better and more information, but have control over the flow of that information and resources for its processing.

This phenomena of detail being pushed up, rather than down the chain-of-command parallels the cognitive deficiencies seen in Autistic patients. A phenomena called fixation or overshadowing takes place when severely Autistic patients attempt to solve problems. Typically they have extreme trouble putting individual problem elements into a coherent picture – a gestalt.

Intelligence, measured by IQ,  amongst individuals is strongly correlated with the ability to move quickly between individual facts and wholistic gestalts, between observations and theories, between axiomata and theorems. This is called attentional scaling. It could be said that the intelligent organization must also demonstrate this capacity.

The filtration mechanisms needed by politicians, or in fact any cognizers, must be those that assist in cognitive scaling – too much theoretical thinking leads to fixation, which exhibits itself in politics as ideological dogmatism and damagogy aka “Principles before facts.”

Conversely, an over-fixation on the individual facts prevents a critical component of problem-solving, insight. That is, the moment of seeing things anew – without generating theories or endorsing them, we can’t hope to learn from experience, individual or collective.

So the answer isn’t simply less vs. more information, it’s a question of the principles for information selection, relevance realization isn’t a process, it’s a constraint on the multiple processes required to cognize. It’s a complex of rules that constrain what information is processed such that the right information gets through, limiting bias and improving insight.

The problem is not that these rules don’t exist – they do, politicans have process information as long as they’ve made decisions. The issue is that the rules for what is appropriate information, for relevance have changed. A system that could realize relevant information thirty years ago can’t cope with the existing informational environment, its fitness has been lost, to use Darwinian concepts.

What system can President Obama employ to ensure strong relevance realization?

One way to assist in the solution of ill-defined problems is the retention of previous solutions for analogous problems, such as old technology.

However, old technology, especially technology that has been wholly eclipsed by modern innovation is quickly become eroded – we are forgetting our inventive past.

This is not a new problem, despite the accelerating pace of invention – the Dark Ages are defined by this collective amnesia, Europeans forgot how to repeat the feats of engineering performed routinely by the Romans – roads, aqueducts (specifically keystone arches) and most forms of sanitation and hygiene disappeared as those responsible for remembering how to do those things disappeared.

The story of the printing press is similar – history’s most famous Alsatian, Gutenberg was the inventor of movable type, not the printing press. Chinese inventors and several different cultures in the Arabian peninsula experimented with printing but found it intractable as a mechanism for archiving, and so the technology died out.

A more contemporaneous example is the recovery of satellite imagery collected prior to the Apollo missions. A cadre of enthusiast scientists have converted a former fast-food restaurant into a photograph recovery lab.  These photos are stored as analogue data on reams of tape. Only a handful of machines exist that are capable of reading these tapes – several had to be cannibalized to get a single one into working order.

Not only is the technology itself endangered, the knowledge of its inner-workings was also rare, only a handful of NASA scientists are still around with sufficient know-how to repair and operate the tape-reading machines. The tapes being recovered retain some of the most detailed images of the moon ever captured – moreover, they capture images of the moon prior to humanity landing there.

This case and another case of a scientist re-purposing a childhood toy (Shrinky Dinks) to create useful and cheap bio-medical tools are documented in the most recent issue of MIT Technology Review.

The process of invention is considered by many to be the introduction of existing (often well-established) ideas into a novel context and exploring its implications.  If we do not have a collective memory and respect of and for past solutions, how often are we spending weeks in the lab to save a day in the library?

Even highly creative solutions like MindTouch or the swirling maelstrom of open source wiki solutions on the web provide different ways to manage content.

While the problem of where to put documents has been answered – technology has offered little in the way of assisting the management of the process of work and creativity. That is, CMS/CRM and the like offer methods to reposit and organize work outputs and lack a coherent and effective system for augmenting the work process, or assisting in the work inputs, particularly inputs exogenous to the organization.

For example, while a CRM (customer relationship manager) application can assist sales and support in tracking data associated with accounts and even provide a way of sign-posting the sales process, most do little in the way of augmenting the client research phase, pitch and messaging development, campaign deployment, data analysis, goal setting and the like. Typically this is done on a whiteboard in a meeting, in Microsoft Word with an implicit process in the mind of the sales manager or simply tracked individually by the sales team with informal activity synchronization over lunch or at regular meetings – hardly taking advantage of the real-time capabilities afforded by contemporary participatory technology.

Worse still are tools for handling work-inputs; the information gathered by workers to inform their work. Particularly, there’s little in the way of process enabling systems to facilitate the process of data -> information -> knowledge. That is, there are few systems that explicitly track the mechanisms, assumptions and decisions by which incoming data is connected with action. In most situations, one must read through acres of discussion threads (either on an intranet or, more likely, through email) such systems are simultaneously fragmented and extremely voluminous. To learn from this information, you must play it back in real-time, a massive productivity burden. Further, it’s often impossible to determine whether a given decision is a relevant precedent or not because organizations typically do not collate and annotate archives of conversation (how pedantic and weird would that be?).

Organizations that DO concretize the process of decision making (rather than merely the results) and store that process for easy retrieval do so implicitly or accidentally – industrial design firms, through their process of prototyping, review and specification are, in many ways, companies that specialize in making particular kinds of informed decisions on the behalf of others – so by professional convention, this process of knowledge management, in the true sense, is built-in. Many organizations, typically lack such a system, or if it does exist is mired in bureaucracy to the point of negating the benefit.

Learning is ultimately what these systems hope to facilitate, learning is in the technical sense, the process of perception that changes behaviour. Without behavioural change, it can be said that no learning has occurred. Which at an organizational level means that only that information which might alter the established momentum of the organization can result in learning. Such a system constraint diminishes the distortion of the confirmation bias in information analysis – by deliberately seeking only alternatives to assumed action, the process of learning will be more robust. The challenge in facilitating learning through software lies in processes conversion, discovery, translation and reconversion.

Conversion

Initially information must be converted from knowledge – someone must commit to writing (typically) what they have learned. This process will invariably produce a sub-set of what was actually learned and so the goal must be to capture as great a portion of knowledge as can be hoped for. Included in this conversion task must be a process for articulating the provenance of the knowledge, the context(s) to which it was/is applied and presumed requisite knowledge to comprehend the information. In systems that facilitate this kind of knowledge conversion, a great deal is made of discovery meta-data. This kind of a priori information architecture management tasks is problematic, as it takes on an erroneous and potentially harmful categorical view of information and knowledge. That such exists in one or even a plurality of enumerated domains. Information is slotted into a taxonomy – either explicitly through some established taxa (say, Library of Congress system) or implicitly through tagging (which in aggregate serve to categorize content in a so-called folksonomy).

This presents the problem as the content author or curator must predict the relevance of the information prior to its employment – an impossible task. The web enabled the feasibility of weak statements of relevance (tags) which is modestly better than strong statements of relevance (explicit categorization) but is still problematic because of its static predictive nature. Solutions like wikis make the process of re-purposing information low-cost, which certainly increases the chances of relevant knowledge discovery but do not address the issues of facilitated conversion – in a wiki you’re still presented with a blank form with no indication as to what to write (that is left as an exercise to the writer).

What this means is that only those people with the conjunction of domain knowledge and the ability to write good prose with little assistance are able to contribute meaningfully to knowledge bases.  The ability to write well is a rare facility and requiring it of every contributor is obviously an issue, further depending on such an ability for the knowledge to be usable is also an issue. KM systems must provide some way for users to capture qualitative and quantitative knowledge without assumptions as to future relevance or requiring the user to compose the structure at every entry.

Conversion from tacit to explicit knowledge can be a worthwhile exercise for improving the knowledge of the users as it can clarify their thinking on a subject as they commit their knowledge back into information. However, the conversion process must facilitate the further steps in the KM process and keep the barriers to entry low so contributors are incented to contribute.

Conversion is the first step, the others are critically important also and are the subject of future posts.

Unfortunately for many organizations such as NGOs this is a disasterous method for initial structuring. Problems like poverty, gender inequity or climate change are not static entities whose assumptions are constantly true over a period of years. Mandates, as they commonly exist for charities and non-profits – especially small-scale, locally focused ones concretize a functional fixedness in the thinking of the organizational members which destroys creative problem solving. However, firms also suffer from that same calcification through the enforcement of “core competencies” – which are intended to be descriptive of the current capacity of the firm but instead become normative parameters for service or product diversification.

Conversely, some mandates are so vague they provide no way to distinguish between competing solutions. Specifically, the mandates don’t provide any way to contextualize innovation within the culture of the organization – how people do their work, the perceptions of members’ and the established norms at play within the organization. This is especially evident in the deployment of technology within many groups. New, fanciful technology is installed with the intention of “revolutionizing the way the organization works.” These applications are generally left derelict because it does not integrate cleanly within the current system. Moreover, technology that has been entrenched often remains because members are not engaged to discuss how technology could improve their work – such tasks are left for upper-level managers and consultant-experts who are often totally disconnected from the on-the-ground practices – referring instead to irrelevant and dated documentation crafted by a different group of removed, decontextualized technical experts.

Intellectual problems can be distinguished into two categories – insight problems and algorithmic problems. Everyone has an intuitive understanding of the difference – algorithmic problems are those with clear goals, established rules and a prescriptive process for their completion. Arithmetic problems are the stereotype for this category, but there are others, such as how to fix the problem of a flat car battery, changing a lightbulb or quench your thirst under mundane circumstances.

Insight problem solving is the hallmark of what William James called sagacity – wisdom, rather than knowledge. These problems are what some would call “non-linear” or “lateral.” The solutions are unobvious at first, but, once described are generally easy to understand by those who understand the problem.  The solutions are famous, often to the point of being folkloric – Newton’s concussive fruit or Archimedes’ wet bathroom floor for example. The psychologist Karl Duncker conducted experiments on this kind of problem, famously he presented subjects with a box of candles, a box of tacks and a box of matches.  Subjects were asked to find a way to affix the candle to the wall, light it and prevent wax from dripping onto the table.  This required subjects to circumvent the “functional fixedness” of one of the boxes as a container and use it as a shelf by affixing it to the wall with tacks.

Importantly, Duncker and others have done hundreds of variations on this experiment, of relevance is the role of incentives. External incentives to complete the task diminished performance – offering a reward made people take longer to complete the task relative to those not offered a reward.

So, external incentives erode performance in insight problem solving, as does the phenomena of functional fixedness.  This is true of individual people, what relevance does it have for groups?

Traditional organizations attempt to solve the problem of group cognition by centralizing it. Recruitment discovers talented, intelligent people and sticks them up top – they become the “brains of the operation” with the remainder of management relegated to autonomic nervous functioning – keeping the operation smooth until signals from the central nervous system change that behaviour. This worked well when the environment in which organizations operated moved at speeds with which it could cope.  Humans have nervous systems attuned over evolution to the speed of operation of the physical world – we have no equipment, biologically speaking, to deal with, say, the speeds of computers.

Traditional organizations implicitly took an approach to cognition that resembles the approach taken by Good Old-Fashioned Artificial Intelligence (GOFAI) researchers for modelling cognition. Simply describe the problem domain, generate a theoretical solution, connect the dots. To describe the problem domain, simply recall all information about similar problems, identify past solutions and transform as necessary for context. The problem domain must simply be defined comprehensively and formally in an interperable fashion to the organization/computer. This was common-sense in the 1950s amongst AI researchers like Hillary Putnam or Jerry Fodor and is standard operating procedure for corporations, governments and NGOs today (crack open an organizational theory textbook and the painful, noble failure of GOFAI haunts the pages).

It didn’t work – we got avocado-coloured appliances but they couldn’t think. AI researchers then believed that computing power would solve the problem – having the ability to search that problem and solution space and connect those dots lay it simply doing everything faster. This sounds like many business-management fads, “lean, lateral teams” or the marketing buzz of acronym-laden business software CRM,CMS,ERP,BPM,OLAP and the conventional assumptions of strategy (react faster to the world around you) and the role of the web (get your word out faster, understand the statistics faster).  Everyone wants to get their inputs faster, believing that a rapid-fire iterative observe-react cycle will solve the problem.

Xenon Pylyshyn, Daniel Dennett and Jerry Fodor can tell us why this thinking is folly – the underlying structures are broken, speeding up non-cognition will no render you cognition – though it might render a sufficient facsimile that you don’t pursue the real solution. The missing component, the real “secret sauce” of this kind of issue is relevance realization. Insightful people are able to put their finger on what is relevant in a given circumstance quickly and easily – they don’t explore the combinatorially explosive solution space of all possible vectors, they go straight for the goods.

To be successful, organizations must do the same thing, they must have processes in place to identify the relevant factors that will guide the formulation of the solution – no individual can do this because no individual can be expected to comprehend the information required. Even if one isolates only the relevant information for a given problem, for an organization of any scale, this exceeds the human capacity to comprehend – we’ve broken the biological barrier to cognition. Thus, the key to an adaptive, sustainable organization is system for collaborative, distributed insight problem solving through a system structure of relevance realization.

How to establish such as system, and the principles guiding its elaboration, are the subject of a future post.

What’s the purpose of OrgCog?

OrgCog will serve as a place for me to craft my thinking about the issue of how groups of people coordinate their actions to accomplish certain behaviour. It will include discussions about knowledge management, computing, organizational theory, politics, cognitive science, communications and the like. I hope to examine ideas surrounding how and why things happen within organizations,  what the current research is on the subject and perhaps discover or illuminate solutions.

Sounds technical.

While some technical information will be reposited, I think in large part the purpose of the blog is to provide an outlet for me to explain these concepts in simple terms – to test my knowledge through communication and in that way solidify my own understanding of the domain.

So, is this a hobby or a job?

Right now, given that I’m an academic, it will be largely a hobby – though I hope to convert the research I uncover here and the ideas I investigate in this blog into projects that may one day expand into money-making ventures – while this is a desired outcome it is not the primary purpose of this blog or my investigations, ultimately I hope that the things I write about in here will be of interest and more importantly useful to those in a position to make changes happen.