The treatment versus experimentation dilemma in dose-finding studies