Algorithms as Evolving Entities: Lessons from Dog Domestication

Among evolutionary biologists there is a theory that pets such as dogs evolved from wild canids because only those that acquired social intelligence could survive. Thousands of years ago wolf packs roamed around human settlements, gradually becoming familiar with human intentions and moods. In other words, their brains began to adapt to human brain activity. Over time their behavior and even appearance became less fierce, more attuned to human emotions, and more symbiotic, eventually turning them into dogs.

This example of dog evolution is used to illustrate that humanity is currently co‑existing with another species that, compared to canids, is more dangerous and powerful: algorithms. The content on Facebook, Amazon, Spotify, Netflix, and even the temperature set by a home thermostat is decided by algorithms. Whenever we interact with the digital world, we interact with algorithms. We need to ensure that these code systems understand our needs and intentions so that we can design products that can feel human and be humane.

Algorithm evolution is part of human evolution

Technology writer Christopher Steiner describes an algorithm as “a massive decision tree composed of a series of binary judgments… a sequence of instructions executed to obtain an ideal result. Information processed by a known algorithm yields the required answer.”

Unlike a domesticated dog, an algorithm’s survival state is not a natural one, and algorithms are human inventions. Yet, like early‑stage dogs, humans still cannot fully understand them, and most algorithms are not designed to respond in ways that match human habits. Algorithms that interact with humans (likely all systems we use, such as stock‑trading platforms) should evolve to be not only effective but also comprehensible.

One point that cannot be ignored in dog domestication is that humans also evolved to live with them. Dogs became part of the human ecosystem. Evidence shows that dogs and humans jointly drove the evolution of brain processes, influencing chemicals such as serotonin. Given enough time, algorithms may similarly affect us, changing the way we think. (Unlike dogs, algorithms will not alter us at the genetic level, but they are already changing our behavior.)

What algorithms do best

Algorithms excel at five things: rapid execution of repetitive tasks, logical decision‑making among alternatives, analysis and prediction, evaluation of historical data, and uncovering overlooked elements. All of these are areas where humans are comparatively weak.

If your work competes with algorithms—for example, high‑frequency stock trading—you are likely to lose. Algorithmic speed is unattainable for humans; even the slowest decision by an algorithm is faster than any human response, operating in milliseconds, akin to a hummingbird’s perception of time. High‑frequency trading already exploits this speed, and exchanges in New York and Chicago will soon achieve near‑light‑speed connections of about 15 ms round‑trip, a speed only algorithms can utilize.

This rapid processing enables algorithms to make judgments across different decisions. These decisions are usually based on data‑driven logical analysis—certain condition sets tend to lead to specific outcomes. The predictions are not always correct, but because algorithms can process massive amounts of data at incredible speed, they can make forecasts and act on results far quicker than humans.

Algorithms also excel at evaluating past events and historical data sets to improve future predictions and provide actionable recommendations. Today we generate massive amounts of data—from large‑scale systems to personal devices and self‑quantification tools. We need algorithms to help us understand what this data means and where its value lies.

All these advantages become disadvantages when humans interact with algorithms, as the same characteristics can lead to friction.

Awkward algorithm interaction

Algorithms can create a new, bewildering experience. In their most effective moments they feel like magic: users receive perfect recommendations or the fastest route home. It feels as if a powerful force—what some call the “Genie Reaction”—is assisting you.

On the other hand, there is the “FAIL Frustration” that arises from algorithmic stupidity—results that ignore context. Algorithms often lack or cannot parse contextual information, leading to situations like a navigation system that doesn’t know about an accident, or a set‑top box that mistakenly recommends content based on misinterpreted audience profiles.

Beyond the quality of recommendations, co‑existing with algorithms produces odd scenarios. In the ending of *Star Wars: A New Hope*, Luke disables the targeting computer. Similarly, we can choose to trust our own feelings and decide whether to rely on algorithmic help. “Beating the Algorithm” can become a fun pastime, even if it sometimes brings annoyance. What if the algorithm’s suggestion is perfect? What if an alternative route is actually faster?

Algorithms can produce uncomfortable, inhuman situations. A route that looks reasonable on a map may force you through three congested streets—possible but barely tolerable, and most people would never choose it. Few people want to be test subjects for algorithmic experiments, yet such cases still occur.

There is also a “value rift”: an algorithm’s priorities may not align with a person’s. Navigation algorithms might claim to save you a minute but instead send you on a winding side street, ignoring your familiarity with the area and the inconvenience of multiple turns. Sometimes a few extra minutes are not worth the trade‑off, but making an algorithm understand that is impossible.

Algorithms: the odd ones among the crowd

Ian Bogost wrote in *Alien Phenomenology* that we do not need to look to other planets for aliens; they already live among us in the form of algorithms. Algorithms are not human; they do not care about or respond to human intentions and emotions unless they evolve—like ancient wolves—to meet human needs.

Unlike wolves, algorithms have not had millennia to evolve. The problems and consequences of rapid algorithmic development are serious. The 2010 “Flash Crash” in the stock market, caused by algorithms, saw the Dow Jones drop 1,000 points in minutes. Imagine a similar failure in the power grid or autonomous vehicles.

Evolving alone

One way to accelerate this evolution is to tell algorithms what human needs and values are. We need to embed awareness of human limits and intentions into code. For example, if a user has never driven a particular route, the algorithm should prefer main roads; if a user appears anxious, it should offer fewer options. When a decision is wrong, the algorithm should be informed—e.g., “this is not the music I like.”

Algorithmic feedback must be adjusted to human cognitive capacity. We cannot input information at code speed. We do not need to know every data point, only the meaningful ones. Telling an algorithm that there is an accident 20 miles ahead may not help, but the algorithm’s calculation could still affect traffic flow.

These code‑based oddities, these ghosts in machines, are becoming more astonishing than their creators. As algorithms begin to take over critical systems, humanity must ensure that, like dogs, they can understand us. Only then might we come to view algorithms as our best friends.