The proposal for cognitive technologies starts with a few assumptions. First, we have goals that we want to achieve in the world. In order to achieve these goals, we need to perceive the world, both to understand its current state and to monitor the progress of our actions/plans. We also need to interact with the world. Second, on an evolutionary timescale, selective pressures should optimize cognitive representations and processes to be useful for achieving our goals. As a result, we should have cognitive instincts.
A cognitive instinct is a predisposition that we are born with, to look for and make sense of certain information in the environment compared to other information. Similarly, we have the capabilities and predisposition from birth to take certain actions to achieve our goals. Importantly, a cognitive instinct does not preclude learning. Proposals for cognitive instincts are often coupled with the idea that we have domain-specialized processing for these types of input, including acquisition processes. Often, components of the acquisition processes are thought to be innate.
Now, there are a lot of rich structures in the environment that are important for us to perceive and interact with (e.g., food cycles, geographical landmarks). There are also a lot of structures that humans have built while achieving their goals and, then, left in the environment. For example, we draw maps and write books that help us navigate and store information. We also put structure into the environment every time we speak out loud (granted this kind of by-product is ephemeral). In effect, we significantly change the environment for future generations. The goals we want to achieve in our environments also rapidly change.
So the environment has rapidly changed (partially due to humans). We now have incredibly different demands of it now within the course of a lifetime but arguably in the past over only a handful of generations. For some of our current goals, we could not have selected via evolutionary processes for mental representations and processes that are required to achieve them. The proposal is that, instead, we selected for domain general learning mechanisms and social learning mechanisms, which we use to construct cognitive technologies—i.e., mental representations and processes that have evolved over generations, not within our genes, but through cultural evolution, cognitive development, social learning and the cognitive artifacts (e.g., speech, maps, books) that humans have left in the environment.
You really should read the Heyes reading.