We want you to help us pick some Oscar winners!

Hey folks, I’m posting this with Tom’s permission. I’m hoping that this would be something that would interest a decent number of people who post here or listen to the Quarter To Three Movie podcast.

I work for a company called Unanimous AI. We do cool stuff to scientifically measure and utilize an idea that’s very near to my heart: that a group of smart humans thinking together can form a hive mind and make smart decisions and predictions and express sentiment in ways that outstrip anything a computer-based AI can do. Because it mimics the behavior (and intelligence) of swarms of bees, we call it “Swarm AI”.

We like to use it to predict stuff, from sports events to stuff like the Time Person of the Year. One of our favorite prediction experiments involves trying to use our tech platform to predict the winners of the major Academy Awards categories each year.

And here’s where you folks come in. We typically contract with a recruitment firm that finds us people who self-reveal as movie fans who are interested in some of the movies involved in the Oscars. But we don’t really know for sure. And so each year we try to find a group of folks who are smart and willing to put some effort into doing a good job of making smart predictions. This year it would be cool if it was our QT3 community!

Here’s how it works:

  1. You take a survey that shouldn’t take more than 5-10 minutes to fill out, in which you make predictions for 16 Oscar categories. We’ll tell you in advance what the categories are.

  2. Then in a live “everyone at the same time” thing, we hope to get 15-20 of you to join together as a “swarm” on our platform. We’ll ask the exact same questions from the survey, and try to get smart answers from a group of folks thinking together as an emergent AI.

  3. Finally, we’ll score our results and share them. And the person here who does the best gets a $50 Amazon gift card. The person here who does second-best gets a $25 card. The person who does 3rd-best get a $10 card.

  4. In addition, I’ve got a bunch of steam codes for some really good games to give away. I’ll give those away to folks just for participating and showing up.

So, some possible questions and answers:

1. Do I need to be an expert? How the heck do I know what the Academy is thinking?!? First, no, you don’t need to be an expert. I figure if you hang out here, you’re probably pretty smart, and if you frequent the movie forum, you like, you know, movies.

But heck, even if you’re a real movie buff, we don’t expect you to be able to figure out just by eyeballing things whether one actor is better than some others in a given year. One of the reasons we’re incentivizing this with prizes is to hopefully get you to maybe check around a little on the web before you register your own personal predictions and then join a swarm. Maybe that means you see what Variety or The Hollywood Reporter is predicting for the Oscars this year. Maybe that means you check some oddsmaker betting sites to see what’s up there. Maybe you spin past fivethirtyeight or something. Any of those are super helpful, because we’ve noticed that even when people do 10 minutes of “scouting”, they improve their own results and our swarm results considerably.

2. So…what do you guys get out of this? What’s the catch? There really isn’t one. In the last 18 months we’ve pivoted away from any social media leanings we may have ever had, so we really aren’t trying to recruit “users” or anything hinky like that. We now split our research between pure science knowledge-gathering to further our own understanding of what it is we can do with swarm AI, as well as work with third-party clients for their own research.

The Academy Awards thing is in the pure research, knowledge for knowledge’s sake category. We’re going to be doing “tracking” swarms on the Oscars right up until they’re handed out, and comparing those results. And, we’d also like some results from some baseline smart people to compare with. Y’all would be our baseline smart people. We’re basically giving away some money and some games to get you folks to help us out. What we need is just to find a time when you folks can give us a half hour to 45 minutes to get together as a swarm.

3. When would you be doing this? That’s a great question. I don’t know! I’m open. My goal is to get at least 15-20 people willing to join us for this. If that means a Saturday or Sunday, cool. If that means a midweek evening after 9 or 10pm EST, great. Whatever works. What would be helpful is to hear from folks interested about what times might work best. We’re about two weeks or so away from the Oscars, so there’s time, but I’d like to do this in that final week before they’re handed out, because that’s when the most research is out there in media and readily available.

Hit me with questions, concerns, and whatnot!

I’d be happy to do it if a time is organised that also suits an Australian timezone. If that ends up being too much hassle then you can drop me no worries.

This is what it looks like when we get a group together as a swarm to answer questions. Each magnet is someone sitting at home somewhere in the world, dragging the clear puck to an answer pole.


I’ve participated in the past, and would gladly do so again. I’ll hit up the survey tomorrow.

As for timing? Any weekday works, and Saturday mornings. Sunday? Not so much.

We don’t have a survey link yet. :)

In fact, I’m writing the survey for our first tracking swarm tomorrow, right now.

We want people to have access to the same info when they take the survey and swarm, so for this project we’ll aim for next week.

I’m certainly up for this.

triggercut, I could never work out what you did for a living. At one point, I was sure you were involved in the record industry, but then I became convinced you worked as a political analyst. Turns out your real job was just as exciting!

Could you go in to a little more detail about how the live portion of the survey will work?

Yeah, my job is a blast! I get to work on predicting things like cryptocurrency and the stock market and NBA basketball and NHL hockey and MLB baseball…and other stuff too. But that predictive stuff is great because it’s fun as heck for everyone involved, but underneath all that we’re looking at some serious hard science that we apply to things like machine learning.

So like I mentioned, there’s two general parts to this.

First, you fill out a survey to capture your individual predictions. We do that, because that survey helps us to generate two scores for comparative purposes: the individual average score of all participants, and then also a “most popular answer” score, where we look to see which answer had the plurality or majority (depending on the question) across all surveys. That’s a sort of “Wisdom of crowds” metric. And then we can measure how the swarm scores versus average individual scores and that most popular answer.

For the surveys, generally we ask that they be filled out some time close-ish to the swarm, like within 24 hours.

Then the fun part: the live swarm activity. We find a time that’s generally agree-able to get us the most participation. You visit our site as a guest, and enter in a username you’d like to use. (So let’s say you use “MrTibbs”; if “MrTibbs” happens to be taken, it’ll pass you through with that guest username anyway, appending it with some random alphanumeric character).

Then you get dropped into the swarm room with everyone else here. There’s a chat area so we can yank one another’s respective chains. ;) I’ll have the Oscar questions pre-scripted, with answers ready to display around that hexagonal diagram. Every participant hovers their mouse cursor near that round air hockey puck-looking thing, which kind of behaves like a planchette on a ouija board. I’ll ask a question, and everyone works together to pull the puck to an answer.

There are lots of cool things going on under the hood. For one thing, the closer your magnet is to the puck, the more pull you have. So engaged users who have a lot of conviction may be constantly moving their mouse cursor/magnet until we land on an answer, as the puck moves. But it also creates a closed feedback loop. You may be seriously wanting one answer, and realize quickly that the answer you want hasn’t got much of a chance based on the way the puck is moving and the magnets arrayed against you. So then you have to decide to maybe switch to your second-favorite choice. Or even third-favorite. It’s basically the opposite of private polling. We’re not looking for individualism and outliers, we’re looking for that magic point where consensus and compromise happen and we get an answer that folks are kind of/mostly happy with.

Thank you for the breakdown. I need to see that puck in action! Watching those individual choices slowly turning into a singular answer in real-time would be a cool thing to experience, and I would love to see how many of my initial picks change from what the group ultimately selects during the swarm event. I would be more than happy to participate.

Consider me interested.

Do you want US residents only? Norway is, as your glorious leader has said, not a shithole country, so I offer my Nordic perspectives on this.

Also, I promise not to answer “Gloomhaven” in all the categories.

Does the live swarm activity happen to have a ‘spectator mode’? I am very interested in this experiment and would love to see it in action, but I literally only see one movie in the cinema each year (Star Wars…), so even after some research I still wouldn’t know enough about the topic to contribute anything meaningful…

Or perhaps you could stream it, afterwards?

The interest is definitely helping me game out when we can do this, so thanks for that, and keep it coming!

Jorn, we’d definitely love to have you from Norway if the times work out, and the same for Tim in Australia.

The timing might be problematic, maybe I failed at reading but what is the designated time.

I am always down for these, despite having seen four movies last year, most of which probably aren’t even nominated for anything.

So what do people think of a week from tomorrow, on Thursday March 1st? Like at 9pm or 10pm ET?

I will most likely be good for 10pm ET March 1st, 9pm ET only maybe.

I would probably be available to share my complete lack of insight (but @triggercut is well-used to the stupidity of my opinions by now)

Hm, won’t work for me, sorry!