The term "risks" is a neologism that came into use with the transition from traditional to modern society. In the Middle Ages the term "risicum" was used in highly specific contexts, above all sea trade and its ensuing legal problems in cases of losses or damages. We find in the vernacular langauges and the printing press of the 16th century the concepts of "rischio" and "riezgo". The context became enlarged to include also life and career at princely courts and other settings where Fortuna plays its fatal role. The English "risk" seems to be a term imported from Continental Europe and appeared only since the 17th century. During these times people thought in terms of good or bad fortune rather than of risk. An increasing "risk awareness" apparently became neutralized by attributing possible future successes and damages to an external source, the goddess Fortuna. This explained, albeit in a metaphorical way, the unpredictability of events and the unforeseeable choice between good and bad outcomes. External attribution protected the decision maker from responsibility in the field of future contingencies. Prudence was viewed as the capacity of humans (as distinct from animals to choose between reasonable expectations, contingent on the actions of other people. So prudence, not risk, was the term for the capability to cope with temporal and social contingencies.

Why, then, was a new term introduced when external attribution could be expressed as "fortune" and internal attribution as "prudence" and when the theory of the logical foundations of cognition had a long and extensive discussion (going back to Aristotle) de futuris contingentibus? Moreover, the term "hazard" or its equivalents, referring to both external and internal attribution of reasonable or not so reasonable behavior under conditions of uncertainty, was already available. So once again, why was a new word without any root in European languages invented? It might have been an evolutionary advance that only later found its final function. Still, there must have been some transitional conditions of plausibility to motivate the use of a new word. Perhaps, this was simply the loss of plausibility of the old rhetorics of Fortuna as an allegorical figure of religious content and of prudentia as a (noble) virtue in the emerging commercial society. Or it might have been the quest for a concept able to accomodate the concrete problems of individual decision making.

The concept of risk, however, remains unclear even today. Normally risk is defined by its antonym "safety" and by its relation to practical affairs. But this amounts to a paradoxical or at least ambiguous definition because in practical affairs there is no absolute safety. The future always contains elements of uncertainty (otherwise it would not be visible as the future, distinguished from the present), and the increasing complexity of perspectives and information may render our knowledge of the future even more uncertain. In spite of a lot of risk research in recent decades this conceptual issue has not been settled. The dominant research fields are ..

All this can be done without a theoretically founded notion of risk but it seems to be generally accepted that there are no guarantees for safety but only risk retreats to safety; and this is particularly so when one takes opportunity cost into account. Risk, then, becomes a universal aspect of decision making. It may even be, as far as future outcomes are concerned, the very essence of a decision. However, the limits of these research fields make it possible to avoid the risk of defining the concept of risk.

Cultural Anthropology and Political Science have contributed the point that "institutions" may be more important than personal opinions for explaining risky behaviour. But this only introduces another undefined term, namely institution, and does not connect with other theoretical contexts. Sociology, too, has developed only a very limited interest in questions of risk. Ulrich Beck has introduced the term Riskogesellschaft but a second look shows that this refers mainly to technological risks. This was a consequence of catastrophes such as Bhopal, Harrisburg or Chernobyl. But there are also risks in capital investment and in speculation in financial markets. There are risks in career decisions, in dating and mating, in "unsafe" sexual practices, and even risks of not getting the needed research funds for risk research after spending much time and effort to elaborate on the research proposals. If it is true that risk is a universal aspect of decision making, we would need a theory of modern society that explains why we have to make so many decisions and why modern society differs in this respect from traditional societies.

If risk is an aspect of decision making, the concept of risk can be defined as the result of an attribution process. It is a construction of an observer. When future possible damages are seen as being caused by a decision, this decision runs a risk - whether or not the negative outcome in fact occurs and whether or not the decision maker takes such a possibility into account or whether it is attributed to him only after the event. (One cannot, of course, avoid risks by ignoring the problem.) Since decisions are always operations of a system, be it an individual person or a social system, we can characterize risk by the internal attribution of possible harm. A possible harm caused by the environment has to be externally attributed and can be called danger. We thus replace by antonym substitution the distinction risk/safety with the distinction risk/danger. And as always, antonym substitution gives a different meaning to the term that appraently remains the same.

This reformulation takes into account that there is no safe way to make decisions. If there are no uncertainities, you simply do not need a decision. But this reformulation also leads into new problems that are, I think, more rewarding to pursue.

First, causal attribution is just a special case of observing within the frame of a distinction (in this case, the distinction of causes and effects), and the designation of something as risk or danger is a form of the second order observing, that is, of observing observers in their situations. This means that causal terms and terms like risk or danger are not indications of ontological facts about which one can have only true or false opinions. The binary logic does not apply. Risk evaluation is not a simply a problem of avoiding error. The question rather is: who uses which frame to guide his observations; and then, who observes how others handle causal distinctions and how they discriminate external and internal attribution depending upon whether they themselves or others make the decisions.

And second, if risk perception is observer-dependent we are able to recognize situations in which risky decisions of one system become a danger for other systems. Both systems can attribute risks and dangers to themselves and to others, depending on the position with respect to the decision making process. The whole problem thereby becomes an internal problem of modern society. Consensus or dissensus in attributing causes and effects replaces consensus about reality and this dissolves all kinds of "reasonable standards" and other normative critera. I shall return to the political and ethical consequences of such a situation in the final section.

With these attempts at conceptual clarification we slide into very complex issues. Yet, this is the only way to adequately explain the relations between the increasing concerns about risks and the structures that characterize modern society as distinct from its traditional predecessors.

The first question is why have decisions became so important in our modern society. To answer this question we need again a detour through unclear conceptual grounds. For it is not sufficient to conceive of a decision as a free (although constrained) choice between alternatives. This usual definition will focus our attention on problems of criteria, values, preferences or conditions of rational choice. These are by no means trivial questions but to define decisions as choice is almost a tautological solution to the problem that has to interest us in risk research. Taking time into consideration, we shoudl rather say that a decision happens when past states and future states no longer match automatically. That decisions have to be "made" is already an interpretation of such a situation (and this leads again to the question: who is the observer?). Such matching can be seen in two ways, as either continuous or discontinuous. If continuous, things, forms, essences stay as they always have been. If discontinuous, we see a process of change such as getting old, growth or decay according to nature. (In this sense, the main description of early modern society of the 16th century was one of "decay".) This traditional semantics of continuity and change was rich enough to accomodate almost all observable states and events because it operated within the frame of a complete distinction, including continuity and discontinuity. It was a distinction of perfect continence. Even decisions could be interpreted within this frame. They produced discontinuity but in accordance with natural ends. And natural ends were described as good ends which, however, would be misconceived by error (Aristotle) or by sin (Christian theology).

It is difficult to know why and how European society could ever have left this closed semantical frame. We find signs of change from the 17th century, for example an increasing interest in newness (as distinguished from deviance); a distinction of ends, and motives or interests "behind" the ends; a new subjective interpretation of security (securitas); a partial replacement of the good (le bien) by pleasure (plaistir); and , of course, the more general use of the term risk and the beginnings of probability calculation. However, a general change of the meaning of almost all traditional social and political terms is observable only during the second half of the 18th century, and even today decision is described as choice and risk as unsafe practice.

My hypothesis is that in early modern times the emergence of new autonomous, self-contained function systems such as state bureaucracy and market economy but also arts and sciences uprooted the traditional aristocratic society based on rank differentiation. This development resulted eventually in a change of the dominant form of differentiation. Hierarchy was replaced by functional differentiation. This change of form was not a "revolution", although the possibility of purely political revolutions was one of its results. It was, in systems theoretical terms, a catastrophe because it replaced one principle of stability, i.e. one form of differentiation, by another one.

As a consequence the temporal structures of the societal system changed and the semantics of time descriptions had to adapt as well. The rhythm of the market cycles is one thing, the planning horizons of firms are already another thing. Technological developments need their time and it is difficult to foresee whether or not they will succeed "in time" - this is so in the case of development of alternative sources of energy production on which society will depend a few decades from now. The time of political elections has another rhythm. How long an individual can keep his job or how long he can survive in a hospital may be uncertain. Religion is not supposed to change at all so that the new fundamentalistic movements and the emergence of new cults have come as a surprise. The implications of changes in one function system for others, for example of religious fundamentalism for the political system or of technological innovations for the markets, are difficult to predict. There is more occasion for surprise and more need for adapting to or evading decisions than in any of the traditional societies. And all this becomes visible within the lifetime of individuals.

At the semantic level, the European tradition had described time as the measure of movements, and this meant that there was one measure, one chronology, for all movements. Chronology provided for the generalization of time, for a universal temporal frame. But this depended upon the distinction between movable and unmovable states, upon the distinction of "tempus" and "aeternitas" [4] and upon the conception of God as the unmoved mover.

This process-related motion of time was used from Aristotle to Hegel and beyond. Even today, it seems difficult to avoid it, particularly for historians. One may replace the distinction of moved/unmoved - with the distinction of fast and slow changes, but then, what does change mean if there are no unchangeable states? Around 1800 a reformulation started. In particular, after the French Revolution and its controversial evaluation, the present was seen as a transitional period, "pregnant with the future", but uncertain in its outcomes. The literary genre of "utopias" was transferred from space to time. The present was no longer the position in time in which one could think of God and salvation. It became the "differential" of the past and the future (novalis); it lost its capability to "present" or "represent" the world and, to some extent, eternity within time, and it became the position of an observer, distinguishing his present past and his present future. With a further moved toward second order observing one can observe in one's present past presents and future presents, that is, past observers and future observers within their time. Time became reflective, it became capable of re-entering itself. Like the "instant" in classical platonic and stoic philosophy, the present became atopon, that is, a position without a place in the world, but as such it became more important than ever. It could now be described as the "blind spot" of an observer that allows him to see time, to use the temporal frame of before and after or past and future to sever the world according to the two sides of a distinction. The form of time is now nothing but this distinction of before and thereafter made by an observer who cannot locate himself within this distinction, who cannot observe before or after his observing.

This condition of an observer who cannot see, but only frame, his observing has an ambiguous (almost transcendental) ontological status. Since Kant we know that the condition of the possibility of experience cannot be based in experience. But whereas Kant thought of transcendental conditions as the foundation of cognition we have to admit that conditions have lost their founding power. They are no longer the ground; they are simply the condition for access to further operations and the condition for second order observing, for observing other observers. And when we search for a post-metaphysical basis of time, being, and knowledge we have to ask what kind of stability can be produced by systems at the level of second order observing.

This is not the place to discuss the far-reaching consequences of this semantic change for a diagnosis of our time, that is, of the time of the societal system with which we have to live. We have to return to our special topic, the meaning of risk and decision in our society. It is now easy to see that "decision" is the answer to the problem of a time with and without a present, depending on whether we observe a decision maker within his present and all his risks or whether we decide ourselves, eventually using self-observation to arrive at the level of second order observing and to see ourselves as conditioning our own decisions.

But what is a decision so that it can be attributed to a system (or to itself?) and not to its environment?

Decisions are events. They are events that happen to the deciding system (again: to persons or to social systems) and to observers in its environment. Strictly speaking, decisions are not made, but they are attributed to an intention, either by the deciding system itself that is aware of making a decision or by its observers. Therefore, omission also can be attributed as a decision if it is clear enough what should have been done. And again, who is the observer who knows, or believes to know, what should have been done?

These attributions may, but need not, converge. And even if attributions seem to converge, their sameness may break down after the decision, if troubles appear. In particular, when the risk becomes obvious and unforeseen things happen, the decision maker may need to defend or even deny his decision, he may look for a sharing of responsibility by others. He may complain about not being given the right information at the right time. He may be forced by events into a position of post-decisional regret. All this makes it difficult to maintain the notion that a decision is a certain type of fact (distinguishable from other facts) that either exists or does not exist and about which one may have correct or erroneous opinions.

The classical definition of decision as (good or bad) choice may be a subjective or an objective definition but it does not take into account the time frame of the decision. Its function was to open the concept for the distinction between good and bad choices and to concentrate the attention on the problem of criteria, goals, values. But "choice" is nothing but a tautological reformulation of the concept. Accepting this definition we may have access to information about subjective preferences, about the need for information and its practical limits and other contextual variables. But we would not be encouraged to ask in what sense a decision distinguishes, and in fact severs, its own past and its own future. And we shall miss the aspect that may explain why the structural changes in modern society that make time so important produce events that appear as decisions, whether we like "freedom of choice" or not.

Decisions reverse the process model of time. Seen as part of a process, the past determines the present. As a result of a chain of past events the present is as it is. It has to be accepted because one cannot undo the past. The future, on the other hand, is open, uncertain and even unpredictable to the extent that it is not simply a prolongation of the past. Decisions, however, reverse this model. They try to find alternatives in the present - as if the past had not simply produced states but also contingencies and therefore possibilities of choice. Moreover, decisions try to give a structure ot the future. They cannot determine the future state of the world or the system but they can project a difference into its open horizons. The standard terminology for this goal, objective, end, aim, but it will be much more revealing to focus on the difference between a preferred state and what would be the case without intervention. Goals need motives, evaluations, justifications that hope for consensus. Differences, on the other hand, provoke second order observer and curiosity about frames and distinctions.

Decisions, then, determine their own identity as events of a special type by recursive relations to a self-designed past and a self-designed future. We can reformulate this by distinguishing distinctions. [5] The past is reconstructed by its memory function, discriminating between forgetting and remembering. The most important function of memory is the repression of past events, that is, forgetting what is no longer relevant, extinguishing its traces and thereby liberating the capacities of the system for new operations. [6] But this repression itself has to some extent repressed to retain, condensate and generalize identities that may be useful to connect the past and the future.

The future, on the other side of the time distinction, is reconstructed by an oscillator function fixing possibilities of bifurcation. If, and only if, a distinction is accepted one can cross from one side to the other, for instance from eating to talking and again from talking to eating. This applies also to the distinction between system and environment or to the temporal distinction between the past and the future. The perspective of (future) oscillations gives uncertainty a specific form. It does not make the future predictable but it has the function of coping with unpredictable events. It replaces the unknown with a binary distinction (a war can be either successful or unsuccessful, may be short or long, but certainly not black or blue). This helps to transform irritation in information as time goes on and to maintain or even enlarge possibilities of decision. One may switch from attention to goals or to attention and means, one may oscillate between labour costs and capital costs; one may, of course, also distinguish distinctions and cross back and forth between scientifically true/false and morally good/bad. Unforeseen results of decisions can be welcome or unwelcome, given certain preferences and non-preferences. The openness of the future allows for seeing both sides of this (and any) distinction, but the frame has to be a distinction. The decision has to select specific distinctions for its oscillator function. In this sense it constructs its future. But it cannot determine its future, because the future retreats as one tries to approach. [7] However, it is possible to fix, for the time being, the frame, or the form, by which one presents the future in the present.

Now we see how decisions operate. They combine memory functions and oscillator functions in highly selective ways. They integrate thereby their past and their future without presupposing this integration as given by nature or by creation. They do not implement a well-ordered world. They perform a contingent handling of contingencies, and only secondarily may they be described in terms of good or bad, right or wrong, safe or risky. Binary codes of this type are secondary reductions that miss the essence (or rather the non-essence) of decisions.

In this sense decisions are always new. [8] They introduce newness, new pasts and new futures, in an unqualifiable world that nevertheless remains the same. The old European tradition thought of newness as deviance. In early modern times newness became slowly admitted as pleasure. Then it became the performance and merit of the "subject". But now we seem to approach a time in which newness becomes the unavoidable utility of fate and risk.

The revision of the concepts of risk, time and decision by referring to background assumptions such as cognitive constructivism, post-ontological world descriptions and second order observation may help us to understand specific features of the modern society. It is obvious that many aspects of life, once taken for granted in continuity or change, are now matters of decision, whereas at the same time the future has become more uncertain than ever because it depends upon the decisions of others. Awareness of risks pervades all sectors of life. The risks of certain technologies on which the society depends are obvious, are present in the mass media and are therefore a topic of political and social concern. But this is only one special group of cases among many others. The complete destruction of the traditional family economies - not only in the landed aristocracy but also in the lower class handicraft and agricultural settings - has mobilized an enormous amount of individuals and of monetary assets. The integration of individuals and society (integration conceived of as a reciprocal restriction of degrees of freedom available in systems) is no longer warranted by family origin and status but by career. One has to make decisions about school careers, professional careers, careers in organizations or careers in prestige and reputation (that is, being mentioned in the mass media), and the question becomes how much effort to invest in an uncertain future when the career also depends upon concomitant decisions of others. Furthermore, the risk in investment in production industries, requiring an ability to foresee the market for many years ahead, is of such import that it seems much safer to speculate in the international financial market because here you can correct your decisions within the next hour or trust your bank to do this for you. As a consequence, those who want to keep their property and remain owners of "their" plants will probably lose their money whereas those who are prepared to buy and sell have at least the opportunity to maintain or increase their fortune. Whether or not enter into intimate relations, whether or not to marry, to divorce and to remarry has to be decided upon, and similarly the question whether or not to have children and at what time in one's life. And there will be much post-decisional regret, because you have got what you wanted in the first place but it turned out not to be the right thing. To engage in scientific research is a risky affair because the expected results may not come forth: and it seems the safer route for many intellectuals to engage in critique and to discuss texts that are already published. The intellectual market, too, has its trade in derivatives. Lawsuits have always been full of risks, in particular when the results depend upon facts to be proven. Nowadays, the complexity of the law (this is by the way, was a complaint already in late Roman republic that led to the development of a special profession of lawyers) and the acceleration of changes by legislation make outcomes even more unpredictable. The most interesting innovation in the system, however, is the results-oriented jurisprudence (nowadays the dominant legal theory) that makes the forecast of the decision of a court depend on the forecast by the judge of the effects of his decisions. And the judge himself is on safe grounds because his decision is legally valid, whether his forecasts are confirmed by later developments or not.

It would be easy to continue this list but this would simply further confirm our impression. The function systems, their organization and the life contexts of individuals in modern society are very different indeed, but the underlying pattern of risk, time, surprise and decision seems to be similar in all cases. The links between these systems are no longer given, for instance in the form of hierarchy or centre/periphery differentiation; they vary from situation to situation. Decisions, then, can no longer rely on their environments; they have to construct their environments and test their constructions. This also means that the links between the past and the future and the degree to which these endless time horizons can be filled with information are no longer given. This, too, has to be constructed on the spot and the decision is held accountable for this. There is no given pattern for doing the right thing at the right time in the right way [9], although systems operate within recursive networks of decisions that confirm or correct one another. The decision making systems have to rely on self-designed redundancies. Since decisions are observed as contigent selections, the risk of their construction, that is, the risk of their specific ways to link a constructed past to a constructed future, has to be attributed to them. There is no other "thing" or "cause" that could be held responsible. (There are of course always other decisions that could be drawn into a broad range of common responsibilities.) Decision is the form of events forced upon a society that has to renounace all natural links between its past and its future.