At this year's Berlinale, I found myself pitching an idea for a script I'll probably never write to an old friend who is now a leading British producer. I felt a bit guilty: this was supposed to be his one non-working lunch of the festival, and there was me turning all Brutus on him. But he had the grace to sound interested as I enthused about the main character, a beautiful, smart, cynical, high-born woman who mistreats her daughter and steals her best friend's husband.

"Okay," said my friend, reverting to his Griffin Mill persona, "so what are her redeeming features'"

That stopped me in my tracks. The honest answer would have been: "She doesn't have any." Instead, I pushed the pan-fried scallops around my plate and mumbled something about her irrepressible energy and her discovery that she was actually in love with the man who is her latest sexual pawn.

The lunch came a few days after the festival debut of Erick Zonca's first English-language film, Julia. I missed it due to conflicting review duties, but I didn't miss the generally downbeat reaction, which seemed to boil down to: "Tilda Swinton is great - shame about the film." Critics are rarely so populist as to pan a film because the central character is not a nice person, but buyers (who make money out of being populist) have no such qualms, and the post-industry-screening comments were reportedly quite credible. Typical of the sample was a dismissal of Swinton's character being "hateful" and so lacking entertainment value.

Distillation of pure evil

Fast forward to the Oscars. As the dust settled on the awards, plenty of mileage was made out of the fact all four acting prizes went to Europeans. Hardly anybody pointed out this year's other interesting trend: that of the characters they played, three were deeply bad people: ruthless corporate lawyer Karen Crowder in Michael Clayton (another Swinton turn), Daniel Day-Lewis' greed-driven oil magnate Daniel Plainview in M and Javier Bardem's psychotic killer Anton Chigurh, a distillation of pure evil that single-handedly lifted the flawed No County For Old Men on to the all-time classics shelf.

True, two of those characters are antagonists - and the antagonist is allowed (indeed expected) to be morally reprehensible. But this still leaves the problem of Daniel Plainview, the 'hero' of There Will Be Blood. Paul Thomas Anderson may be an established name, but even he had to pitch his adaptation of Upton Sinclair's 1927 novel Oil! How would he have answered the question about his protagonist's redeeming features'

By stressing, perhaps, Plainview's fascinating, monomaniacal energy and drive; by mentioning the potentially sympathetic contrast between his business acumen and his gaucheness; by mooting that, with their country again entangled in a foreign war that has turned sour, US audiences were finally ready for the return of a protagonist who exposes the grubby side of the American Dream - like those classic 1970s anti-heroes Harry Callahan and Travis Bickle. I like to think, though, that Anderson would have given more of a Plainview answer: "This film isn't about redeeming features."

Supporting the hero

It's empowering to resist the script-seminar doctrine about sympathetic heroes (Robert McKee: "The hero must have that one positive trait through which the audience participates in his fate"), just as it's empowering to resist the emphasis on the primacy of conflict in good dramatic structure. Where's the conflict in Juno or Into The Wild' Any self-respecting script guru could find one ("It's in the refusal of two young people to conform to social norms and expectations"). But the standard protagonist-antagonist character clash is absent: both films are packed with good, caring people who support, or at least do not to hinder, the hero's goal.

So, is bad is the new good, and support the new conflict' Maybe. What does seem clear is that the sort of audience responses that scriptwriting doctrines, producers' greenlights and buyers' chequebooks depend on are more fluid and less easy to second-guess today than they have been for quite a while. Until a definite pattern emerges, taking risks could be the new playing safe.