The most important thing to consider is not how many cards each team still needs to win, but how many turns they need. You should be more willing to take risks in cases where success will decrease the number of turns you need while failure will not decrease the number of turns your opponent needs. For example, assume for a moment that other than the current turn, a team can always get 2 cards per turn, and on the current turn you can either get 1 safely or 2 with a high chance of giving your opponent 1. Then if you each have 4 cards left, you should take the risk (your opponent needs 2 turns no matter what you do, but you'll need 3 turns if you don't take the risk), while if you each have 5 left then you should not take the risk (you need 3 turns no matter what you do, but your opponent will only need 2 if you give them a card).
Now in reality, it's not as simple as getting 2 cards per turn - if you and your opponent are at similar skill levels then I suggest actually pretending you are in your opponent's chair and try to figure out clues for their team as well in order to tell how many turns they need.
(For your specific example, it sounds like you need 2 turns no matter what because by assumption any two-word clue you give will run your team into 'painting'. If you give your opponent 'painting' then they definitely need only 1 turn while if you play it safe then they may need 2 turns, so you should play it safe.)