An Auditable Complexity Ladder for Mechanism-Level Experiments in Artificial Life

In the vast field of artificial life, scientists build digital organisms to see how complexity arises from simple rules. They are searching for a phenomenon called open-ended evolution, a state where a system never stops creating new forms, new behaviors, and new solutions, much like the endless innovation seen in Earth's natural history. For decades, researchers have struggled to prove they have achieved this. Many have claimed to see it, but their methods often lack a clear way to verify the results or rule out accidental tricks. The community has agreed on what open-ended evolution looks like when it happens—novelty, diversity, and increasing complexity—but they have not agreed on how to measure the specific mechanisms that cause it, or how to prove that a result is real and not just a fluke of the computer code. Without a standard way to check these claims, the field remains full of assertions that are difficult to trust or repeat.

Huazhang Shen, an independent researcher, has proposed a new way to solve this problem. Rather than claiming to have built a perfect open-ended system, the researcher has built a rigorous testing framework, a "complexity ladder" designed to make every claim about artificial life auditable and transparent. The work centers on a minimal digital cell, a tiny computer program that mimics the basic parts of a living thing: a boundary that holds it together, a way to eat and maintain energy, a set of instructions to copy itself, and the ability to divide. The researcher added features to this cell one by one, creating eleven distinct levels of complexity. At each step, they set strict rules before running the experiment, defining exactly what success would look like and what would count as a failure. The goal was not just to see if the cell got more complex, but to prove that each new ability was caused by the specific change made, and not by some hidden error in the code.

The process relies on a method called "freezing" the criteria. Before a single run of the experiment begins, the researcher writes down the specific numbers that will determine if a level is successful. For example, if a new feature is supposed to help the cell survive stress, the exact survival rate required is written down in advance. Once the experiment runs, the computer checks the results against these pre-written numbers. If the numbers match, the level is marked as confirmed. If they do not, the result is rejected. This prevents the researcher from changing the goalposts after seeing the data. To ensure the results are trustworthy, the system uses a "byte-safe" isolation method. This means that if a specific feature is turned off, the computer code changes in a way that is mathematically guaranteed to be identical to the state before that feature existed. It is a digital guarantee that turning off a mechanism changes nothing else in the system, allowing for a clean comparison between "with" and "without."

The ladder starts with the most basic requirements: a container that holds the cell together and a way for it to maintain its energy. The first few levels confirm that without a boundary, the cell cannot persist, and without metabolism, it cannot divide. As the researcher adds layers, the cell gains the ability to copy its instructions, link those instructions to its physical traits, and eventually regulate its own growth. One of the most significant findings occurred at the level of membrane self-replication. The experiment showed that the cell's outer boundary must be able to grow and divide on its own for the cell to truly function. Without this ability, the cell barely divides, and natural selection cannot act on it. This proved that the boundary is not just a container but a causal gate that unlocks the potential for evolution.

Further up the ladder, the cell learns to sense its environment and communicate with others. When the digital cells could sense a signal and respond with a delay, they survived better than those that reacted instantly. When they could share information to coordinate their actions, they thrived in crowded conditions. The researcher also tested what happens when cells specialize. In one experiment, some cells were programmed to stop reproducing and instead produce a shared resource for the group. This division of labor only worked when the shared resource actually provided a benefit; if the resource was useless, the group of specialized cells performed worse than a group of identical, self-reproducing cells. This confirmed that cooperation is only advantageous when it solves a real problem.

The final levels of the ladder tested the cell's ability to adapt to new, harsh conditions. The researcher introduced a stressor that the cells had never seen before. The cells that could mutate their instructions to find a solution survived, while those that could not died out. However, the researcher found that this rescue only happened when the stress was strong enough to kill the non-mutating cells, but not so strong that even the mutators could not survive. An initial analysis had suggested a "stress window" where evolution worked at lower intensities, but the researcher caught that this claim was based on a flawed comparison where conditions did not match. After re-running the experiment with a corrected design, the result showed that no detectable advantage existed below a specific intensity threshold; the evolutionary rescue effect only began once the stress reached magnitude 6, not lower. The researcher also tested whether a cell's ability to maintain a stable internal state (homeostasis) helped it evolve. Surprisingly, the experiment found no evidence that being stable made the cell better at evolving; the two abilities were separate.

Crucially, the paper does not hide its mistakes. The researcher openly reported several times when the initial design of an experiment was flawed. In one case, a claim about a "stress window" was based on a comparison that did not actually match the conditions being tested. The researcher caught this error, re-ran the experiment with a better design, and published the corrected, narrower result. In another instance, a test designed to prove that a specific part of the cell's code was responsible for survival failed because the test itself caused too many cells to die. Instead of ignoring this, the researcher reported the failure, explaining that the specific cause of the survival advantage remains an open question. This willingness to disclose negative results and self-correction is presented as a core part of the method, ensuring that the final picture is honest.

The study concludes that while the researcher has not yet created a system that evolves endlessly, they have created a trustworthy template for how to test such systems. The eleven levels of the ladder serve as a proof of concept for a new standard of rigor in artificial life. By specifying criteria in advance, isolating variables with mathematical certainty, and reporting failures as openly as successes, the work provides a way for the scientific community to verify claims about evolution. The paper establishes that the discipline of testing is just as important as the discovery itself, offering a clear, auditable path forward for anyone trying to understand how life, or something like it, can begin and grow.