Evolution is a complicated thing. Much of modern evolutionary biology seeks to reconcile the seeming randomness of the forces behind the process — how mutations occur, for example — with the fundamental principles that apply across the biosphere. Generations of biologists have hoped to comprehend evolution’s rhyme and reason enough to be able to predict how it happens.

But while prediction remains a worthy goal, scientists are now focusing on its much more ambitious cousin: control over how it happens.

This may sound like science fiction, but the greatest examples of the endeavor live in our past. Consider the process of artificial selection, a term coined by Charles Darwin: Thousands of years ago, humans began to identify plants and animals with preferable traits and selectively breed them, which amplified these traits in their offspring. This approach gave us agriculture, one of the most transformative cultural inventions in human history. Later, artificial selection in animals and plants helped us understand genetics, and how genes evolve in populations. But as effective as it’s been, artificial selection is still fairly limited.

This is different from natural selection, the force that drives adaptive evolution on Earth, where there is no intentional actor doing the selecting. The selecting actor is not a human breeder, but nature itself, which selects the variants with the highest “fitness” — those with the greatest likelihood of surviving and producing healthy offspring. And when nature does the selecting, the outcomes can be difficult to predict.

Now biologists hope to dictate how evolution happens at the molecular level, and to exert as much direct control over the reproductive process as we do in crops. Can we orchestrate evolution, mutation by mutation, toward whatever outcome we prefer? 

Remarkably, we’re already partway there. The 2018 Nobel Prize in Chemistry recognized work on a method called directed evolution, which allows scientists to engineer new biomolecules. One of the winners, Frances Arnold, pioneered a way to mutate proteins in the laboratory and then measure their functionality — say, how well an enzyme metabolizes sugar. It’s then possible to isolate the protein candidates of interest, mutate them, and select further, until we have generated a protein with improved function (in this case, an enzyme that metabolizes sugar very efficiently). In this sense, chemists are operating like dog breeders, but without relying on sexual reproduction to generate the protein offspring. Rather, they are generating a diverse population of proteins and measuring their properties in mere hours. And by selecting what they want, they are controlling how evolution happens.

From this example, it is becoming clear that controlling evolution — steering it toward certain outcomes — requires knowledge of how evolution will happen, along with the technology to intervene. So we can think of the problem through the lens of a simple equation: Control = prediction + engineering.

This control can be more subtle than Arnold’s approach. One 2015 study suggested using antibiotics in a certain order to steer evolution away from creating antibiotic-resistant pathogens. And something similar is happening with cancer treatment: Oncologists are trying to leverage our molecular understanding of cancer to steer cancer cells toward susceptibility to certain drugs. This is possible because we know that when a cancer cell evolves resistance to one drug, it might become more susceptible to others. This notion of “collateral sensitivity” is based on fundamental principles of trade-offs in biological systems: In general, there is no “free lunch” in evolution, and adaptation often comes with costs.

In more recent work, scientists have generalized these approaches. Using ideas from quantum physics, a multidisciplinary team (including physicians, computer scientists and physicists) applied a method called counterdiabatic driving to shift populations toward predetermined goals. For example, infections from some strains of malaria parasites are easier to treat than others. Researchers might try to “drive” populations of the parasites toward the more easily treatable strains.

Similar ideas are being applied to other systems, such as the microbiome, where evolutionary biologists are now using directed evolution to control microbial communities like the ones that live on our skin and in our gut. To do this, they are using knowledge of how certain microbes interact with each other along with new microbial techniques that allow us to introduce microbes into a population of other microbes. The hope is that we can use this knowledge to one day steer the composition of the microbiome to one associated with improved health outcomes.

These breakthroughs demonstrate that in some form, evolutionary control is a thing of the present, not the future. But most successful examples have taken place in a small number of settings: microbes, microbial communities and proteins. And even further, existing efforts focus on control over short time periods — no reasonable scientist purports to be able to control molecular evolution acting over decades or centuries (outside of the artificial selection that has taken place over millennia). True control over the evolutionary process remains strictly limited by our current knowledge and tools.

While the technical challenges of evolutionary control remain substantial, the ethical barriers are also notable. The issues overlap with those around genetically modified organisms. When we engineer a mutation into a strain of corn that confers the ability to grow even in stressful environments, we influence future generations of that strain of corn. Furthermore, embryo selection in humans can resemble artificial selection, giving us the ability to steer the appearance of human traits in future populations. In general, overzealous applications of these technologies can be driven by a kind of genetic determinism — the naïve view that the meaningful differences between organisms within a population can be explained (mostly) by their genetic makeup.

Should we ever try to naïvely steer evolution in humans and other organisms over a longer timescale, we would fall victim to a sort of evolutionary determinism, which holds that we can and should have full control over how life evolves in the future. Ultimately, these ambitions are misplaced. They underestimate the caprice of biological evolution — the difficulty of considering all the forces that shape how life functions and flourishes. Some may imagine that artificial intelligence can help resolve these uncertainties. But AI is not a panacea for ignorance. It is most effective when we already understand the vagaries of the system that we are attempting to model and predict. Evolutionary biology doesn’t quite meet this standard — at least not yet.

We can (and should) simultaneously gush at the ambition of modern biology and have the presence of mind to recognize our limits. For example, the eugenics movement suggested that the human race could be improved using the sorts of methods that gave us domesticated animals and crops. We now understand it was both bigoted and based on bad biology. Examples like these are cautionary tales, and they should teach us that careless attempts to control tempestuous forces like evolution are bound to fail.

Quanta is conducting a series of surveys to better serve our audience. Take our biology reader survey and you will be entered to win free Quanta merchandise.