At various times I have been part of dinner groups that meet once a month or once a week at either the same restaurant or a different restaurant every time. There's usually no special arrangement, but it's usually good for the restaurant since they get a big crowd on a slow night.

I think there could be ways to make it better for the restaurant as well as the diners -- and the rest of the web to boot. I'm imagining an application that coordinates these dinners with diners and the restaurants. The restaurants (especially newer ones) would offer some incentives to the diners, plus some kick back to the web site for organizing it. As part of the deal, the diners would agree to fairly review the restaurant -- at first on public restaurant review sites and/or their own blogs, but with time at a special site just for this purpose. Diners would need to review at least 80% of the time to stay in.

Here's what could be offered the diners:

• Private rooms or private waiter, with special attention
• Special menus with special items at reduced prices
• Special billing, either separate bills or even pay online -- no worrying about settling.
• Advanced online ordering and planning for shared meals, possibly just before heading out.

For the restaurant there's a lot:

• A bunch of predictable diners on a slow night
• If they order from a special menu, it can be easier and cheaper to prepare multiple orders of the same dish.
• Billing assistance from the web site with online payment
• A way to get trustable online reviews to bring in business -- if the reviews are good.

Now normally a serious restaurant critic would not feel it appropriate to have the restaurant know they are being reviewed. In such cases they will not get typical service and be able to properly review it. However, this can be mitigated a lot if all the restaurants are aware of what's going on, and if the reviews are comparative. In this case the restaurants are being compared by how they do at their best, rather than for a random diner. The latter is better, but the former is also meaningful. And of course it would be clear to readers that this is what went on.

In particular, I believe the reviewers should not simply give stars or numerical ratings to restaurants. They can do that, but mainly they should just place the restaurants in a ranking with the other restaurants they have scored, once they have done a certain minimal number. This fixes "star inflation." With most online review sites, you don't know if a 5-star rating is from somebody who gives everything 4 or 5 stars, or if it's the only 5-star rating the reviewer ever gave. All these are averaged together.

In addition, the existing online review sites have self-selected reviewers, which is to say people who rate a restaurant only because they feel motivated to do so. Such results can be wildly inaccurate.

Finally, it is widely suspected that some fraction of the reviews on online sites are biased, placed there by the restaurant or friends of the restaurant. There are certainly few mechanisms to stop this at the sites I have seen. Certainly if you see a restaurant with just a few high ratings you don't know what to think.

This dining system, with the requirement that everybody review, eliminates a good chunk of the self-selection. Members would need to review whether they felt the mood or not. (You could not stop them from not going to a restaurant that does not appeal to them, of course, so there is still some self-selection.) It is possible a restaurant might send its friends to dine at "enemy" restaurants via the club to rate them down, but I think the risk of this is much less than the holes in the other systems.

Restaurants with any confidence in their quality should be motivated to invite such an online dining club, especially new restaurants. Indeed, it's not uncommon for new restaurants to offer the general public things like 2nd entree free or other discounts to get the public in, with no review bonus. If the site becomes popular, in fact, it might become the case that a new restaurant that doesn't invite the amateur critics could be suspect, unwilling to risk a bad place in their rankings.