It is a very fundamental principle indeed that Knowledge is always gained by the orderly loss of information, that is, by condensing and abstracting and indexing the great buzzing confusion of information that comes from the world around us into a form which we can appreciate and comprehend. (K. Boulding)

How do humans acquire knowledge? It is often associated with empirical reductionism, which fits the worldview of Modernity. It is important to understand this method and its shortcomings, find ways to evaluate the quality of the resulting ‘scientific statements’ and to explore other ways of acquiring knowledge – or Knowledge.

Doing a scientific project

Suppose you are concerned about air pollution and set up an experiment to measure the concentration of substance X in a well-defined area. The measuring tool is itself an illustration of scientific development. The result of your experiment is a series of concentration values at given location p and time t, c(p,t). Building upon atmospheric physics and chemistry, you interpret the results in terms of dynamic cause-and-effect processes with for instance a Gaussian plume model. You then realise that it is actually the impact of air pollution on the neighbouring forest that matters. With the help of ecologists, you extend the model with a forest dynamics model. You do longitudinal experiments (>5 years) to translate exposure to impacts on forest health.

Unfortunately, you cannot rely on such solid laws in this field as in atmospheric science. Estimates of tree sensitivity are based on controlled laboratory experiments and fields surveys, but they are only partly transferable to the field situation. The trees in the forest differ in age and in location-dependent parameters such as soil and water access. There is also interference with other species and other pollutants. Besides, the forest may have varied responses to fluctuating or prolonged exposure, which falls outside your measurement domain. It may take much effort and many years before you have valid scientific knowledge about the impact of air pollution on the forest.

One evening, you meet a friend who argues that the real issue is whether the measured air pollution has a negative effect on the health of the people living in the area. Taking up the challenge, you ask some medical scientists to engage in a longitudinal research project to measure the health situation of people in the area and, for comparison, of people in another area with negligible air pollution. Although there is substantial medical knowledge about how the air pollutant affects the physiology of the human body, your long-term experiment is confronted with large uncertainties. For one thing, the group of people followed in the experiment is not constant because people move in and out of the area. There is also a large variety, both somatic and psychic, in the population samples. For instance, some individuals are more sensitive to exposure, while others are better able to cope with the effects. It is difficult and ambiguous how to deal with this variety. Often, ad-hoc research strategies have to be designed. The experience has surely made you less naïve about ‘scientific knowledge’.

The ordeal is not over yet. At an environmental economics conference, an economist argues that one does not need to know the impacts of the air pollutant concentration in great detail; it is more important to know at what cost it can be reduced below some level which is considered or negotiated as an ‘acceptable’ risk. Recognising the appeal of this argument, you get another research project funded with economists to estimate the options and costs to reduce the concentrations. You calculate the emission reductions needed for a ‘safe’ concentration level and some economist colleagues identify and rank the emission reduction options according to the cost per unit emission reduced. The epidemiologist in the team wonders how to set the ‘acceptable’ or ‘safe’ level of pollution, so an additional cost-benefit analysis is started to explore the trade-off between (healthy) lives saved and the extra costs.

The government wants policy recommendation on the basis of your research. The team discusses the various policies. One economist argues that polluters respond mostly or solely to price incentives so a policy should focus on the proper tax levels. They quote several policy analyses and behavioural surveys to proof their point. Colleagues disagree and quote scientific evidence in favour of emission standards for equipment: policy should enforce stringent standards in order to elicit technological options that reduce emissions at much lower cost. During the deliberations, you discover that some of your colleagues’ relationships with government officials and entrepreneurs seem to play a role in their convictions about the most successful policy measures. Demonstrations by affected farmers and residents put you under pressure. You realize that the dispute cannot be settled in the way of the natural sciences, if only because controlled experiments are excluded. Besides, the parameters involved keep changing all the time.

As this story illustrates, if you widen the system boundary, ever more scientists and stakeholders get involved. Questions and answers become more complex and uncertain. People come into the picture, with their own psychological and socio-cultural characteristics. You as investigator enter the scene, with your skills, limitations and biases. It is time for some philosophical reflection.

Strong and weak knowledge

The physicist Groenewold (1981) has proposed a judgment in terms of strong and weak science. A scientific theory about a particular object system develops by strengthening three elements:

  • logical operations (l);
  • codified experiences (e); and
  • hypotheses (h) which relate the experiences.

The codified or coded experiences refer to human experiences reframed in experiments which are set-up to interact with the natural system and are the outcome of the conversion from percepts to observations. This is the inductive part of science. For instance, enjoying wild deer in a natural park is not codified but a well-documented count of wild deer in a certain area and period is.

The logical operations constitute a more or less formalised system of concepts and rules. They are the rules of the formal system and the language for conversion from conceptual to mathematical model. It can be a language syntax, differential-integral calculus, transition rules in a cellular automata model or decision rules in an agent-based model (§9.4). It is the deductive part of science. For instance, a differential equation to describe the exponential growth of a deer population is a formalisation of a particular set of observations, in particular on animal reproduction. The more it is formalised, the sharper the concepts and rules can be – but also the more the object system is simplified. The hypotheses connect and expand the observations via logical operations. The observation that the number of offspring of deer fluctuates with the number of wolves can induce the hypothesis that the two populations are interacting. The formalised hypotheses yield a scientific model.

A scientific theory becomes mature by gradually eliminating unnecessary hypotheses and sharpening the codifications of experience and the logical operations. When a theory grows out of hypotheses that have been falsified and replaced by hypotheses better in agreement with observations, the resulting statements are called strong(er). Much knowledge in physics and chemistry represents such strong knowledge and have become natural law.

[1] It is formulated in mathematical models that are thoroughly empirically validated within their domain of observation and control. Yet, there is often still much scope for improvement. An example is meteorology, where the combination of satellite data and simulation modelling did and still does lead to improvements in weather forecasting.

A theory is strengthened by eliminating weak knowledge. Sometimes such statements are obvious, as in false logic, misapplied statistics or a priori judgments with a claim on absolute and universal validity. Sometimes their weakness is only obvious if one demands empirical evidence, as in statements about non-observable entities with no consequences for experience or in theological dogma’s or revelations. Sometimes, statements reflect an intuition or speculation, which the Zeitgeist is not yet ready for or not willing to consider for cultural or political reasons. Notice the importance of language in this assessment of the quality and validity of (scientific) statements.

Footnotes

[1] Often the words hard and soft are used to denote what is meant here with strong and weak. The strong-weak terminology is preferred, because hard and soft are better used to refer to the degree to which an aspect of reality can (or cannot) be influence (by men). Calling knowledge in the life and social sciences weak(er) sounds derogatory. This is not the intention: what is meant is that the objects of study in these sciences are too complex for the natural science methods of controlled experiments and mathematical formalizaton.