Imagine that you take the smart watch off your wrist, place it on a banana, an apple or a lemon, which is what I did a few minutes ago. Obviously, the fruit is no longer alive, it does not receive sap, it has no way of showing biological activity. And yet… A heart rate of 65 beats appears in a lemon and 105 in an apple, as seen in the image. In real time.
It seems like a joke. After all, Many fruits have bones, skin… but heart? The reality is that smart watches are not “listening” to the heart, what they do is an estimate based on light.
If we look at the bottom of a smartwatch we see several small green LEDs that flash constantly. Those LEDs are part of a technique called photoplethysmography, or PPG. The idea is simple: Each time the heart beats, a small extra amount of blood reaches the blood vessels near the surface of the skin. Because blood absorbs some of the green light, the watch detects tiny variations in the amount of reflected light.
Thus, more blood means less reflected light and vice versa. By analyzing these oscillations thousands of times per second, the watch calculates your heart rate. The problem: the watch doesn’t know what it’s looking at. It can be a human doll or a fruitthe smartwatch only sees light patterns and if it detects changes sufficiently similar to those produced by a pulse, it will try to interpret them as heartbeats.
And it turns out that some objects have internal structures capable of generating optical signals that confuse the sensor. That is the case with many fruits: They are full of fibers, water channels and plant tissues with complex optical properties.
When LEDs illuminate that structure, some of the light is scattered, absorbed and reflected in ways that a watch interprets as the fluctuations that human blood flow would produce. To this we must add that modern algorithms are designed to “fill in gaps.” If the signal is weak or imperfect, the software attempts to reconstruct it using statistical models. That is, it is not limited to measuring; He also interprets.
The curious thing is that this phenomenon is very similar to what happens with other systems, such as artificial intelligence that is often programmed to detect patterns or radars. A weather radar can mistake a flock of birds for a storm because they both produce similar echoes. Radar sees no rain; sees reflections. Likewise, the watch doesn’t see heartbeats; sees light variations. The “fruit pulse” appears when these variations are sufficiently similar to those of a human being.
Does it mean that watches are unreliable? At all. When worn on a human wrist, today’s watches are typically quite accurate at measuring resting heart rate. What this experiment shows It’s not that they work poorly, but that they work in a different way than we imagine.. They are not little doctors directly observing our heart, but rather optical instruments that observe light patterns and make inferences.
And this should make us understand a little-recognized truth about modern technology: Most of the machines we use every day do not “understand” the world, they recognize patterns and there is a big difference.