Knowledge is of paramount importance to modern societies. Scientific knowledge, in particular, is regarded as indispensable for meeting the great challenges of our time. At the same time, the limits of knowledge or ignorance or non-knowledge and how to deal with them are an important driving force of scientific thought and action.

Laboratory experiments are a central tool of science for testing hypotheses and generating knowledge. In particular, when exploring the interactions of innovations with the environment and society, laboratory experiments reach their limits and the step from the laboratory to exploratory experiments becomes necessary. Such real-world experiments are carried out in particular to test and develop new technologies in a natural environment or in their future places of use. The aim is to create new knowledge or to transform non-knowledge into knowledge. Existing non-knowledge is thus the motivation for conducting an exploratory experiment.

At the suggestion of the Committee on Education, Research and Technology Assessment, the background paper explores how to deal with this dilemma - the dilemma of relying on scientific experimentation to generate knowledge, fill knowledge gaps and drive change and innovation, while at the same time facing the inevitable ignorance of possible undesirable consequences of experimentation on nature and society.

Contents and structure of the background paper

To address this tension, two heuristics - the technology characterisation method and the perspective of recursive learning in real-world experiments - are presented and illustrated using the three research areas of green genetic engineering, fracking and ocean fertilisation with iron as examples. The two approaches studied differ significantly in how aspects of non-knowledge are analysed and assessed, and how this can lead to a better and more responsible way for society to deal with the challenges that may arise from research, development and innovation.

Technology characterisation is a risk and hazard management approach. It combines an engineering and scientific perspective with the precautionary principle and focuses on anticipating potential hazards, so-called reasons for concern. The perspective of recursive learning in real experiments is closer to a social science view of societal learning. It assumes that anticipatory risk research cannot predict all relevant possible effects of exploratory experiments and therefore relies on concomitant observation in the sense of institutionalised learning in the context of knowledge application as early as possible. The approaches differ in their focus and tend to be complementary, i.e. they complement each other and can be combined.

The background paper is preceded by a brief summary. Chapter 2 introduces key terms and principles and discusses different manifestations of scientific experimentation. Chapter 3 discusses basic ways of responding to scientific ignorance, before Chapter 4 introduces the two approaches to dealing with ignorance in exploratory experimentation, the characterisation of technique and the perspective of recursive learning in real-world experimentation. Chapter 5 takes a brief, general look at different forms of social participation. Chapters 6 to 8 then deal with the concrete handling of non-knowledge in exploratory experiments in the three research fields of green genetic engineering (Chapter 6), fracking (Chapter 7), and marine fertilisation (Chapter 8). Chapter 9 summarises the following aspects of good governance in exploratory experimentation.

Aspects of good governance in exploratory experiments.

By its very nature, the subject of non-knowledge remains subject to limited specification and is always the subject of political debate in the case of controversial issues such as large-scale technological exploratory experiments. It is therefore not surprising that non-knowledge is often instrumentalised in political disputes. Questions about how to value non-knowledge in the context of a particular experimental endeavour are therefore at their core political questions, and decisions about whether and under what circumstances to undertake a large-scale exploratory experiment are inherently political decisions.

Elements of how to deal with exploratory experiments.

In the run-up to an experiment, in addition to approval in principle (if required) and funding, decisions need to be made about the design and configuration of the experiment from a technical perspective and in terms of further process support. Issues to be considered include

• Inclusion of all available risk information, further preliminary studies if necessary;
• structural and event-related vulnerability analysis (in the sense of technology characterisation) in addition to an analysis of the intervention;
• for experiments of high concern: a staged approach with upscaling of experimental stages in  space and time;
• Consideration of potential societal benefits as part of a comparative risk assessment;
• Planning of appropriate societal involvement.

During the course of the experiment, other issues will arise that should be anticipated in order to indicate the need for termination, readjustment or mitigation. This can be done through the following structural measures:

• Establishof a process-accompanying, continuous or recursive form of monitoring and (re)steering to respond to new findings; if possible, define termination criteria in advance;
• appropriate design of the monitoring activities, both in terms of targeted monitoring of specific parameters and broad screening for possible surprises; definition of a minimum monitoring period;
• Balanced composition of the accompanying steering committees; allowing and taking into account unsolicited forms of participation;
• Ensuring a high degree of transparency of the data, results, decision criteria and decisions taken.

Combining ex ante analysis with in-process control.

Good management of non-knowledge in exploratory experiments must attempt to reconcile two objectives that are in fact incompatible: To explore the extent of non-knowledge, including "unknown unknowns", as systematically and derivatively as possible, or to characterise it ex ante, and on this basis to take precautionary measures if necessary, which can range from a modified design of the experiment to a moratorium. At the same time, however, the design of the experiment should create space and openness for new insights and surprises, otherwise the experiment as such would not make sense.

The two approaches presented in the background paper, technique characterisation and recursive learning in real experiments, should be seen as heuristics that can be used sequentially to achieve good results, i.e. judgments in the sense mentioned above.

They represent two complementary approaches to the problem of dealing with non-knowledge. It makes sense to collect as much risk knowledge and evidence of cause for concern as possible in advance - this is, among other things, the approach of technology characterisation - and, after (provisional) approval, to additionally establish a continuous monitoring and control process in order to be able to react to newly emerging knowledge, new assessments and also newly emerging ignorance (generally speaking: surprises) - this is the perspective of recursive learning in real-world experiments.

Monitoring and aggregation of data and knowledge

A crucial prerequisite for noticing and registering emerging ignorance and surprises during and after the execution of the experiment is broad and, if necessary, long-term monitoring. A balance has to be struck between narrowly focused, hypothesis-driven observation on the one hand, and attention to surprises in the sense of broad monitoring on the other. In addition to the variety of parameters to be monitored, the duration of the monitoring is crucial. Surprises can occur with a (large) time lag. Here, too, a difficult trade-off must be made between effort and cost and the longest possible observation period.

Since scientific observation programmes alone cannot be expected to capture surprises systematically or completely, openness and attentiveness to observations by different actors, including non-scientific actors, are important. Such an attitude can be implemented, for example, by establishing a body where relevant observations are collected and systematised (such as the pharmacovigilance system in medicine).

Comparative assessment of risks and benefits

The level of risk a society is willing to bear is a question that is constantly being renegotiated politically and socially. The expected benefits to society must also be taken into account. The prerequisite for a well-founded evaluation and decision-making process is scientifically comprehensible information on the risks and benefits that can be reasonably expected.

Societal participation

Societal participation
Large-scale exploratory experiments take place in the social sphere. Crucial to the meaningful design of participation are questions about who is affected: Who has a voice, who should legitimately be heard, have a say, etc.? Even if it seems obvious to answer the question of affectedness as objectively as possible from the perspective of the direct (physical) consequences of an experiment, such an approach is not sufficient because the physical, natural consequences are only partially predictable due to a lack of knowledge. Moreover, the economic and social consequences of a new technology in a complex, networked world are unlimited. The question of "affectedness" is therefore a socio-political question. It is therefore obvious that when designing participation processes, care should be taken to ensure that citizens have a say in whether they feel affected, as in the case of uninvited participation.

In democratically constituted states, uninvited participation is an important building block in opinion-forming and decision-making. It is an expression of perceived civic responsibility and can help to bring aspects into the discussion that would otherwise be neglected or only marginally considered. Uninvited participation can therefore be seen as a socially positive element of a living democracy and can be made fruitful, e.g. by making the positions, assessments and knowledge contributed transparent, and should not be hindered by secrecy.