There’s a phenomenon that I’ve come to call the “Akinator effect.” In 2007, the French company Elokence released the AI-powered game Akinator. It is essentially a digital 20-questions game in which the user thinks of a real or fictional person, object, or animal, and the Akinator guesses who it is by asking a series of yes-or-no questions.

A decade ago, when it was first launched, Akinator was absolutely astounding. It could accurately identify even the most obscure characters from little-known works, usually in fewer than 10 questions. It was nearly impossible to stump the AI within the 20-question limit. I remember it correctly guessing the Captain, a minor character with a half-dozen lines in Act IV of Hamlet, in five questions! Furthermore, it would learn from its “misses” through user feedback, constantly improving its abilities.

However, over time, its performance drastically deteriorated. The AI started to repeat questions – I suspect that the AI thought they were different questions due to similar user-entered questions with minor differences. Furthermore, the quality of its character data declined. Within just a year of gathering user input, the platform transformed from amazing to highly disappointing.

In short, the quality of the output declined because the system allowed users to add their own data, and much of that data was low quality. In other words, noise was added to the system, drowning out the signal.

The Akinator Effect in Business

Far from being confined to the computer world, this effect is everywhere. A good example is data reporting in organizations. Often this reporting starts with simple numbers. And then more and more are added to it until it becomes unwieldy and, often, almost unusable.

For example, imagine a coffee shop that has just started collecting data instead of the owner estimating everything. At first, the data is amazing. It lets her know what items are selling and what the inventory levels are, which helps them ensure things are in stock. This is so useful that they expand to start looking at trends.

They find that Tuesday is a big day for decaffeinated coffee, allowing them to stock up ahead of time. This data is so useful that they start asking for more data. They start looking at how much items cost to buy and how much they sell for. This is also useful.

Pretty soon, they have page after page of detailed information about each product and customer, showing that people whose last names start with “H” spend 6.23% more than the average customer. No one can read this dashboard without training on what all the pieces mean and how they fit together.

In short, what began as a huge benefit quickly became a burden, and a burden that was expensive to create because the coffee shop owner assumed that since a little information was helpful, a lot of information would be even more helpful.

Social Media

Initially, social media platforms used simple algorithms to show users content from their friends and people they followed. This worked well. However, as these platforms evolved, they introduced increasingly complex algorithms to determine what content to show users.

These algorithms now consider hundreds of factors, including user behavior, engagement patterns, content type, and even emotional responses. While intended to increase user engagement, this complexity often leads to echo chambers where users only see content that reinforces their existing views and lots of sensationalized content.

Complaints about difficulty in finding posts from close friends and general frustration with lack of control over what they see abound. The result is that many users find their social media experience less satisfying and less useful for genuine connection, despite (or because of) the advanced algorithms designed to improve their experience.

Cars

A big offender of this is cars. Cars used to be fairly simple affairs with a steering wheel, seats, and just the basics for getting from here to there. Over the years, the additions have added up, and now cars have so many switches, buttons, and knobs that most of us don’t know what most of the buttons in our cars do.

Sure, these extras can be ignored most of the time, but if you are trying to find something you need, such as turning on the air conditioning, all of the extra buttons and knobs make that much more difficult to do. It also added a lot of cost to the car to add these things that are seldom used. It’s a lot of added cost and complexity for little added or even negative value.

Ok. Then why are they there?

This begs the question, “If the value is negative, why even add them in the first place?” The reason is because people erroneously think it will be valuable. Imagine you are going to buy a car. You see two cars that are fairly identical, but one has a navigation system plus Bluetooth and a direct wire connection which also charges your phone, while the other one only has a direct connect for your phone. The price difference is $100.

Even though you like your phone maps application, you think “I might use the navigation or the Bluetooth, and it’s not that much extra, so I should just go ahead and get it”, never thinking about how the extra options might make using the car more confusing. This is probably no big deal with one item but do that with item after item and the complexity of the car and the added costs explode.

You might notice it in the additional cost of the car, but you probably won’t be thinking about how much all those new buttons and knobs will make the car more difficult to operate, and more likely to break. Similarly, the coffee shop owner keeps adding new data points because she thinks each new one might be useful and is only thinking about the costs to show the new data point, not the added difficulty in finding the most important data points this adds or that any of the new data points add to the complexity and potential of the system to fail.

Signal Versus Noise

We’ve all heard the idea of signal versus noise. This comes from telecommunications and refers to the useful information (signal) versus unwanted interference (noise). The problem is that a weak signal can also be noise. Imagine you are in a crowded restaurant with loud conversations everywhere. Your friend is talking to you across the table, and you can’t understand them over all the other voices.

To you, those voices are nothing but noise, but to the people talking, they are signal conveying useful information to their own friends. In the same way, the volume knob for the radio in your car is important if your desire is to change the volume, but nothing but noise that might get in the way if you want to turn on the air conditioning.

This isn’t a real problem when there are only a few items, but when dozens of buttons and knobs proliferate, as they often do in today’s cars, it can be very hard to figure out how to turn on the air conditioning.

Conclusion

More isn’t always better, especially when it comes to information and complexity. Whether in AI systems, business reporting, social media, or product design, the drive to add more features or data can often lead to diminishing returns or even negative outcomes.

The key takeaway is the importance of balance and purposeful design. Instead of constantly adding more, we should focus on curating and refining what’s truly valuable. This means:

1. Prioritizing quality over quantity in data collection and feature implementation.
2. Being willing to stop adding new items!!!
3. Regularly reviewing and pruning unnecessary elements.
4. Focusing on user experience and ease of use.
5. Understanding that simplicity often leads to greater effectiveness and satisfaction.

Be mindful of the Akinator effect. The goal should be to amplify the signal while minimizing the noise, ensuring that additions genuinely enhance value rather than merely increasing complexity.