Scientific journals have a credibility problem. Here's how to fix it