In the past year, I've become disillusioned with computational recommendations. I now believe that they can work, but the computation will only get you halfway there.

I've noticed that given enough data, it's possible for a computer to predict what I like. At the same time, it's hard for me to trust algorithmic recommendations. I don't trust them because I know they're from a computer, and I find it difficult to believe that a computer program can "know" who I am. This is a hundred times worse if I'm the one guilty of writing the program.

Any recommender is trying to predict the future. The recommender is guessing what the receiver will find interesting, or relevant, or compelling enough to consume. This prediction is important, but can only get you so far. As with baseball scores, guessing that the home team will win will make you correct more than half of the time. All additional information will make your prediction better, but prediction is hugely affected by the law of diminishing returns.

That's the reason the most effective recommenders – while making great predictions – work to sway the future in their favor. This is possible because everyone's judgment of what's good is imprecise. People I trust can sway my judgment. This is a good thing. If a friend tells me a wine is of high quality, my brain will make it taste better.

Good recommendations work because you trust the recommender. Without this trust, a recommendation algorithm cannot work.

According to our data, the article recommendations given by Scoopinion work pretty well. On average, people spend four minutes on a story recommended by us – three times as long they spend on an average story found via Facebook. (The articles we recommend are also three times longer.)

Even though our readers realize that an algorithm can't know their heart and soul, they know that the stories unearthed by Scoopinion are of high quality in general. They are selected from the pool of stories that the reading community has spent the most time with. Being a part of the reading community builds additional trust.

I call this trust the placebo half of recommendation. And as we've learned from medicine, trusting the pill is sometimes just as important as the pill itself.