options for a given situation, reflecting further discoveries of institutional and sociological factors that shape technical opportunities (Herdt, 2006; Ojiem et al., 2006). This understanding has deep roots in extension research (e.g., Loomis and Beagle, 1950; Ascroft et al., 1973; Röling et al., 1976), farming systems research (Collinson, 2000), 1980s gender research (e.g., Staudt and Col, 1991; Sachs, 1996), and 1990s policy research (e.g., Jiggins, 1989; Christopolos et al., 2000). However, the reasons that thinking about policy began to change likely had little to do with this research and more to do with the realization that technology supply-push could fuel massive social problems wherever there were no alternative opportunities for those who could not survive in farming. This lack of survival contributed to the growth of megacity slums (UN Habitat, 2007), the ease with which displaced youngsters eagerly turned toward civil disorder and even civil war (Richards, 2002; UNHCR, 2007) and the growing numbers of internal and transboundary migrants (UNHCR, 2007). Supply-push arrangements were shown to produce agricultures accounting for 85% of the world's water withdrawals and 21% (rising to 35%) of gaseous emissions contributing to climate change; and to the declining material condition of natural resources and biophysical functioning (MA, 2005; UNEP, 2005). The cumulative evidence indicated a policy change was overdue.

     The concept of innovation systems offered itself as a policy model for sustaining agricultures to meet ecological and social needs. Effective innovation systems were shown to need systemic engagement among a diversity of actors (Havelock; 1986; Swanson and Peterson, 1989; Röling and Engel, 1991; Bawden and Packham, 1993; Engel and Salomon, 1997; Röling and Wagemakers, 1998; Chema et al., 2003; Hall et al., 2003, 2006). However, people and organizations interact in diverse ways for the purposes of creating innovation for sustainable development; the range of actors needed to develop a specific innovation opportunity is potentially large and thus becomes increasingly difficult to classify (Figure 2-3) (see 2.3). The "innovation systems" concept, widely used in other industries, usefully captures the complexity (Hall et al., 2006) by drawing attention to the totality of actors needed for innovation and growth; consolidating the role of the private sector and the importance of interactions within a sector; and emphasizing the outcomes of technology and knowledge generation and adoption rather than the strengthening of research systems and their outputs.

     Empirical studies emphasize that the dominant activity in the process is working with and reworking the stock of knowledge (Arnold and Bell, 2001) in a social process that is realized in collaborative effort to generate individual and collective learning in support of an explicit goal. Innovation processes focus on the creation of products and technologies through ad hoc transformations in locally specific individual or collective knowledge processes. As such innovation is neither science nor technology but the emergent property of an action system (Crozier and Friedberg, 1980) in which knowledge actors are entangled. The design of the action system thus is a determinant of the extent to which an innovation meets sustainability and development goals.

2.1.4.2 Market-led innovation

From about the 1990s onwards innovation processes in agriculture principally have been driven by a rise in marketled development. Typical responses to market pressures in North America and Europe in terms of the way in which technical requirements, market actors, and market institutions interact can provide an understanding of the "innovation space" for socially and ecologically sustainable agriculture (NAE Chapter 1; Figure 2-3).

2.1.4.3 Technological risks and costs in a globalizing world

The risk outlook fifty years ago could be described in general terms as high local output instability, relative autonomy of food systems and highly diverse local technology options: an agricultural technology that failed in one part of the world had few consequences for health, hunger or poverty in other regions. The increase in aggregate food output and the trend toward liberalizing markets and globalizing trade has smoothed out much of the instability; it has integrated food systems (mostly to the benefit of poor consumers) and it has spread generic technologies throughout the world for local adaptation. The mechanisms of food aid, local seed banks and other institutional innovations have been put in place to cope with catastrophic loss of entitlements to food or localized production shortfalls. Yet the world faces technological risks in food and agriculture that have potential for widespread harm and whose management requires the mobilization of worldwide effort (Beck et al., 1994; Stiglitz, 2006). A robust conclusion is that human beings are not very good at managing complex systemic interactions (Dörner, 1996).

     Immediate costs of risks that cause harm typically are carried by the poor, the excluded and the environment, for instance with regard to choices of irrigation technologies (Thomas, 1975; Biggs, 1978; Repetto, 1986); crop management (Repetto, 1985; Kenmore, 1987; Loevinsohn, 1987); and natural resource and forestry management (Repetto et al., 1989; Repetto, 1990; Repetto, 1992; Hobden, 1995). The weight of the evidence is that power relations and preanalytic assumptions about how institutions and organizations actually work in a given context shape how scientific information and technologies are developed and used in practice, producing necessarily variable and sometimes damaging effects (Hobart, 1994; Alex and Byerlee, 2001). Recent assessments for instance of the "long shadow" of livestock farming systems (Steinfeld et al., 2006) and of agricultural use of water (Chapagrain and Hoekstra, 2003) lead to a well-founded conclusion that estimations of agricultural technologies' benefits, risks and costs have been in the past too narrowly defined. The mounting scale of risk exposure in agriculture is delineated in the Millennium Ecosystem Assessment (2005), Global Water Assessment (2007), and IPCC reports (2007).The accumulating weight of evidence that past technology choices in agriculture have given rise to unsustainable risks has led to efforts to develop more appropriate technological risk assessment methods (Graham and Wiener, 1995; Jakobson and Dragun, 1996; NRC, 1996) and to take on differing perspectives on what levels of harm are acceptable and for whom (Krimsky and Golding, 1992; Funtowicz and Ravetz, 1993; Funtowicz et al., 1998; Scanlon, 1998; Stagl et al., 2004). Important experience has been