I would define it in two parts:
1. Consciousness of facts: What you "know" is what you are conscious of. Machines are already conscious because they have memory.
2. Meta-consciousness: You know that you know, or you know what you know. This is a feedback loop on knowledge, unlimited in recursion. Sort of like the check if a pointer is null (and a check if the check is null, etc.).
I know hardly anybody agrees with this, but I'm pretty serious about it. People tend to confuse what they see as the world with what any entity should see. To make computer more human, is to create human senses, simulate human techniques of interpretation and memory and add a comparable experience of time.
But in my eyes there is no reason not to claim that computers are already conscious. Consciousness just isn't the goal.