Rational Agents and Emotional Humans: Why you should include variable rewards

It’s one week until John’s birthday, and you just got the an invitation from Lisa for his birthday party. In the good ol’ days, you’d add it to your day planner, make a mental note to get a gift, and check if you had anything nice to wear.

Today you’d might get the invitation on Facebook, and it would show in  your digital calendar. The invitation would include information such as guests, the address to the venue, a map with estimated driving time, and the price of an Uber. You’d still have to remember to get a gift, and dress yourself.

All illustrations are licensed from YAY. Click on illustration to see portfolio.

Soon, everything might be even more streamlined. You might still get your invitation on Facebook, it will appear in your digital calendar, and so on. But, we already have the technology to take this a step further.

The perfect intelligent assistant

What if you Intelligent Assistant, I.A., could suggest gifts for John? It could use his social media activity. Or, a bit creepier, cookie tracking products he’s added to a shopping cart, but never bought. You could get great gift suggestions, and with one click have it wrapped and on it’s way.

You I.A could check the dress code and style for the venue you are going to. A bit creepier? It could check the social profile of other guest, and calculate an appropriate attire. Since we pay almost everything with our credit cards, all your purchases are saved and can be id’ed. This way, your wardrobe could fully digitalized. Then, based on situation and what other people wear, matched to create a great outfit.

Of course, your I.A. would provide you with good small chat topics and talking points. It would just check the interests of the other guests, and the latest news.


Come Saturday, and you are on your way to the party. Since it’s at a bar, your I.A. recommended taking an Uber. Your outfit is on point, and you are confident the gift will be a huge success. Your I.A. just alerted you that there will be at least three single women that might be a good match at the bar. The bartender knows your favorite drink since it’s included in your social profile. She has it ready for you when you walk in, since Uber notified her about your arrival time. The DJ is an I.A., creating a playlist that matches the people, age group, occasion and time of day. (What it won’t know what to do though, is to play that cheesy song your friend played over and over again in 2003.)

What is wrong with this story? (Hint: Perfect is not rewarding)

It’s not that it won’t be possible, most of this is – someone just needs to connect the dots. (And the government would have to not protect people’s privacy).

It’s the simple fact that humans are not perfect, we are not rational, and perfect is boring.

Why does this matter? Let’s look at the tasks the I.A. would do:

  • add to calendar, and remind you
  • find the perfect gift
  • buy the perfect gift
  • assemble the perfect outfit
  • assess all the people attending
  • find great small talk topics
  • navigate to the venue
  • match making
  • favorite drink
  • music played

It’s not only at work – our unique, personal skills are also becoming commodities

I don’t know about you, but in my group of friends and extended network people have different skills, and some skills they value more than others. Eric is always organized, and a life saver when it comes to remember what needs to be done. Mary is mind-blowingly good at finding thoughtful gifts. Richard is always on point with his outfits, and knows how to help others add that extra flare. Emma always has a funny story or an interesting take on current events.


What happens when these traits becomes common? When everyone can master every skill that we used to put a high value on?

Would you compliment a gift if you knew your friend ‘only’ did what her A.I. recommended? According to NYTimes, you would. Though, your friend might not value the compliment as much if there was no effort in the hunt.

How about that drink? Would you try new stuff only when an algorithm suggested that you might like it? How would you discover anything radically different? What about “black swans“?

This touches in on social behavior theory, as well as product design and habit theory.

The importance of variable rewards in product design

Nir Eyal writes about the importance of variable reward. Variable rewards is what makes us come back, to keep using a product.

There are three main forms of variable rewards.

The Hunt – You can hunt for products, information, offers, bargains, rewards.. It can be you browsing a store, physical or online, looking for the perfect gift for John. The reward is the trill when you find it.

The Tribe – Oh, the tribe. Social recognition. Likes, shares, comments – we love to get validation and to feel that we belong. This isn’t a bad thing. Look at all the people spending hours to help others, because they feel part of a ‘tribe’. Wikipedia is one example, Facebook another.

The Self – The joyful feeling of competences, of mastery. It feels great to be good at something, and to get validation for it. How about the feeling when your inbox is empty? Or making it to the next level?


An I.A. can take away some of these good feelings. Will you feel as great if you didn’t have to hunt for anything? Everything just showed up. “Here: the perfect whatever-you-wanted”. Would your weird fun facts be as funny and weird if anyone can find them? “OK Google, tell me a fun fact about dolphins”.  How about all the great map-readers out there? Their competence is becoming a commodity, unless there’s a crisis  – like a dead phone.

An I.A. can replace more and more of our ‘unique’ skills. We are talking about this for the workplace, but it will have an impact on our social lives as well. What we need to remember is to add a human element to our I.A.s, and to our products in general.

Example of  variable rewards in ecommerce

This might change in the future, but looking at it today I don’t think people want perfection. They might say they want it, but perfect is boring. When you are building a product you should look at the Why’s, then think about the emotional aspects and variable rewards.

For our ecommerce solution, soon out in beta, it would be stuff like:

The Hunt – Although you’d pretty fast get good data on preferences, we’d like to add a mix of products. One mix of customer knowledge and rewards could be to show different elements in the product feed. Bargain hunters would get random bargains, socialites would get random exclusive items.


The Tribe – You can share your interior photos, and ask for feedback from friends, family and your network. You can create and share collections, helping others find what great stuff.

The Self – You’ll gain competences in interior decoration. You might get great deals for those special items that makes your living room look yours, and not a copy of Crate and Barrel. You’ll gain rank as others recognize your work, and when you help others.

Our goal is not to create a social network, but to add a social element to our shopping app. Hopefully, customers will enjoy the combination of new technology and emotional rewards.

We’ve only seen the beginning of I.A. and the technology it’s built on. I think the potential it has is beyond what we can grasp. Here’s hoping for a great relationship between humans and I.A.


Read part two:

Rational Agents and Emotional Humans: Why you should ask “Why?”

An intelligent agent is by definition a ration agent, who will “acts so as to achieve the best outcome or, when there is uncertainty, the best expected outcome.” With the agent being rational, and humans emotional – how can we tell a rational agent what the best possible outcome is?

How do you translate emotional needs into a rational goal – when we might not even know the answer ourselves?