Defining conservatism is surprisingly hard; well, perhaps not that surprising, since self-described conservatives have adopted many apparently incompatible positions in recent years. Hayek advocated small government, but Reagan and Bush increased its size. The religious right and extreme libertarians claim the term. Thatcher asserted there was no such thing as society, a proposition which her successors have denied hotly. Deficit hawks demand a balanced budget, but tea partiers refuse to raise taxes under any circumstances. Prime Ministers such as Diefenbaker (and Macmillan in the UK) administered corporatist governments; the Reform Party was populist and neo-liberal. European conservatives love the European Union, which the British loathe.

Does this matter? Isn't this just fighting over a word? It does matter, because words matter. 'Conservative' has over time come to mean 'right wing', which debases our political language. It matters because it allows critics of conservatism to accuse it of incoherence, or to make up their own unflattering characterisations (philosopher Ted Honderich felt able to define it as 'organised selfishness'). And it matters because it names an important and useful political tradition with great potential value for today's complex and dynamic world.

The project of my book Conservatism is to develop a meaningful definition that will demonstrate conservatism's importance, and pose demanding questions for alternative ideologies. I begin with the simplest association of ideas: conservatism should surely be about change. Even that flies in the face of recent history-Thatcher was a radical, Bush thought nothing of turning entire societies upside-down. But I have faith in etymology; I make it my primary axiom.

What properties should my definition have? I make three demands on it. First, it should be based on principles, and not be an enumeration of 'conservative' points of view. My book is therefore not a review of conservative literature, and many great names are absent from it. Second, it should bear a family resemblance to at least some parts of the conservative tradition. Third, it should be couched in the language of public reason-its premises must be meaningful to non-conservatives. That means, for instance, that I could not define conservatism in terms of Christianity (as many thinkers, such as Russell Kirk for example, have done), because those terms are not meaningful to atheists or other non-Christians.

The key concept is uncertainty; the conservative challenges the ground on which rationalist ideologues stand (this argument owes much to Hayek and Oakeshott, continuing a line from the Ancient Greek sceptics through Montaigne). The first principle upon which conservatism is based is therefore the knowledge principle:

Because society and its mediating institutions are highly complex and dynamic with natures that are constantly evolving as they are co-constituted with the individuals who are their members, both data and theories about society are highly uncertain.

That in itself says nothing about change (indeed Hayek welcomed change, and even wrote an essay called 'Why I Am Not a Conservative'), but if we add to the mix a discussion of risk in the spirit of Ulrich Beck, Jared Diamond and Nassim Nicholas Taleb, we get the second half of the definition, the change principle.

Because the current state of society is typically undervalued, and because the effects of social innovations cannot be known fully in advance, then social change (a) must always risk destroying beneficial institutions and norms, and (b) cannot be guaranteed to achieve the aims for which it was implemented. It therefore follows that societies should be risk-averse with respect to social change, and the burden of proof placed on the innovator, not his or her opponents. It also follows that change, when it does come, should ideally be (a) incremental, (b) reversible where possible, and (c) rigorously evaluated before the next incremental step.

From the conjunction of these two principles, many things follow, including well-known conservative themes such as authority and tradition; social stability, the rule of law and sound money; conformity to norms; plurality of values; personal responsibility; opposition to populism, fundamentalism, planning and targets.

Note also that this is not a philosophy that condemns change; a sensible reading of the two principles shows that they merely shift the burden of argument to the innovator. Change is more desirable when the state of society is less satisfactory, and therefore when the risk of change is relatively small. Switching from the language of opposition to change, to the discourse of risk, is vital for conservatism's modern-day relevance.

We should note three interesting wrinkles. First of all, conservatism thus defined is relative to the state of a society, and can therefore be completely different in different places. The US conservative thinks a written constitution essential for freedom; the British conservative thinks the opposite. There is no contradiction; they defer to separate traditions. A conservative from Alberta sounds very different from a colleague from Ontario, or from Quebec. And all three are very distinct from conservatives South of the border.

Indeed, US conservatives, defending a liberal tradition ('liberal' in the John Stuart Mill sense of valuing individual freedom above all), sound very like liberals themselves. This is why, I argue, free market liberalism and even libertarianism have become tangled up with conservatism in recent political thought; the predominance of American conservatives has led to the false assumption that the aim of conservatism is to defend American mores. That is only the aim of American conservatism.

En passant, let me note that neo-conservatism is yet another kettle of fish. Although neocons want to preserve their own societies, they think nothing of disrupting others, because they believe the imposition of their own norms and values will improve other places. This is neither risk-averse, nor opposed to change. It is more akin to old-fashioned imperialism than conservatism properly understood.

The second interesting corollary of the definition of conservatism is that it is not necessarily a right-wing philosophy. A left-winger could deploy conservative arguments to preserve welfare, health or social security systems, for example.

And thirdly, the conservative should not hark back to a golden age. Trying to recreate the past is risky social engineering just as much as trying to create a wonderful rationalist future. Broadly speaking, the conservative should not try to stand in the way of major social movements, but should rather concentrate on mitigating their effects on functioning societies.

Conservatism is highly relevant in today's troubled world. No conservative welcomes debt, and all genuine conservatives should be wary of untrammelled financial innovation (indeed, Canada's staid banking system is an example to us all). Though big government is a bad thing, the conservative should try to reduce its size in incremental stages, not giant disruptive leaps. The current vogue for government transparency is excellent; it will put Edmund Burke's little platoons online, and help ensure social change is demand-driven, not theory-laden.