Matt Dillahunty, host of The Atheist Experience TV show brought up an excellent point to remember when talking to people about this issue. Ask them:
"If you had a better way of acquiring knowledge and truth, how would you demonstrate it?"
Of course, the answer would be to use a scientific approach! In other words, we'd need to use science to prove a better method than science...
Matt also had another great example on the show. He told a caller to pick Heads or Tails. The caller said, "heads", Matt flipped the coin and it came up Heads. There is no way the caller could have known heads would come up but yet, they made a correct prediction - one could say the caller 'knew' it would be heads. However, what benefit is there to studying how the caller knew this? Did the caller 'know'? Did the caller have a currently unexplainable method of 'knowing' how the coin would land? Maybe, but in order to find out, we'd have to test his claims! In order for the caller's claims to knowledge to have any validity, the caller would have to demonstrate that they can predict the coin flip with a far higher frequency than chance on repeated trials. In other words, use the scientific method, to know.