I understand your view point but I'm inclined to think that in the same way that technology is an extension of nature, so is the complex nature of animals,including humans. I can't separate the agency of the animal from the genetic survival mechanism.
How can we really know if the ability of a person to use the abstract or learned behaviour is genuinely separate from a pre programmed instinct ?
I would point to the sometimes unpredictable nature of survival strategies and the apparent sliding scale of protecting genes belonging to ones self, children, cousins and sometimes the community. There seem to be very subtle rules in place that guide these things and they are not always obvious. It might be that the genetic control is so complex in its nature that it seems to be separate from simple biology. Maybe we just can't comprehend fully the reach of the gene and it's control of it's biological vehicle in the same way that that it seems incredible that nature produces highly complex interactions between species using only evolution.

The agency of the animal feeds into the genetic survival of the species, our survival instincts and drives are very strong, and they are improved by the capacity of agency, more than they are failed by it, because they enable us to improve our capacity for flight or fight, through learning strategies and capacity to adapt on the fly.

The selfish gene isn't really controlling peoples behavior per se, but that the way our agency and biological interaction works, improves that capacity to pass on genetic material and the benefits of our agency (sort of like a meme to a gene).

Of course agency is generated on a biological mechanism, through the interaction of numerous systems, and maybe illusionary, but our experience suggests that what it does is serve as a capacity to override or fine tune things stimulus-response.

In terms of a mechanistic system, I don't subscribe to that - Its a modernist concept that breaks down under a post-modern scientific analysis, and a by product of laboratory bias. In reality I believe systems to be gear more towards Chaotic and fuzzy logic, than to a simple stimulus-response mechanism.

Mechanistic models of nature invariable lack a suitable predictive capacity to suggest that they are true, outside of strict controls. I also would argue that all mammals, at least, have a degree of agency and self awareness, along with a capacity to learn and adapt, that agency creates.

If evolution is about 'species best fitted to the environment' agency is an inevitable conclusion of any evolutionary process, as it creates the capacity for a creature to adapt its behavior to fit changes in its environment.

Of course any model of agency, would appear to mechanistic, because in describing the properties of any model, the capacity of the whole is reduced to the function of its parts.

The important thing with chaotic systems is that they are both at the same time chaotic and disordered (neither is dominant) what matters is the point of scale at which you interact.

How we measure something, and the methodology by which we measure something, will inevitably affect our results, especially when dealing with hypothetical concepts (such as in behavior). In psychology we see this a lot, where classical research and empiricism breaks down entirely when critical qualitative methods, such as discursive and phenomenological methods are applied, even with some of the most respected and well known research.

I think that sciences create models of reality, rather than being exact representations. I suspect existence before translation through sensory data closely resembles a more 'chaotic system', and that in order to function we construct meaningful represenations of it, rather than 'experience it' as it truly is.

But that is what the evidence suggest, that the method of observation and measurement scale used, will affect the outcome and results; this presents the idea seen in post-modern science outlooks, that rather than constructing a view of reality, scientific methods construct a model of reality based on our observations and measurements (which actually fits better with experimental vs applied science dichotomy).

Scientifically speaking there is no real basis to assume otherwise, as the chaotic and fuzzy systems present a solution to the macro vs micro scale that fits both levels, and works.

Science isn't about truth, its about predictable and highly probable outcomes. The mistake of modernism was to assume that the scientist is somehow entirely objective and separated from the experimental methodology (and as we experience the world subjectively, we can never truly ever separate our bias entirely).

In truth, science has never really been anything other than the construction of models of high probability over low probability.

As such the Quantum and relative are both acceptable models.

Unlike the clear sighted religious view.
And next, a debate on what colour shoes pixies wear.