Continuing the discussion from New insights on defining "biblical kinds"!:
Once a super-intelligent designer who has the ability to order nature in a fashion outside its observable processes, laws, and boundaries is posited, I would think it impossible to predict in a scientific way what super-intelligent, outside-natural-boundaries ordering actions that designer would have taken.
In his book Darwin’s Black Box, Behe proposed that ID-driven biology would inter alia simply abandon any research that would attempt to explain through drift, mutation, and selection the advent of (allegedly) irreducibly complex structures and systems. To the average scientist, this proposal does not in the least look like science.
But obviously you are coming at the question differently than I. Would you like to explain how in your view upper case Intelligent Design is a scientific theory, Eric?
In addition, evolution makes all sorts of predictions that have led to productive scientific research. For example:
Fossil record: Early life would be single-celled, then small multicellular forms would appear, then larger and more diverse forms. Importantly, as more chronologically intermediate fossils are found, phenotypes intermediate between earlier and later species should have some statistical likelihood of appearing. Neil Shubin used this prediction to target late Devonian geological formations to search for intermediate forms between early Devonian lobe-finned fishes and Carboniferous amphibians. And he discovered Tiktaalik.
Nested hierarchy from common ancestry. This is a prediction of evolution and it does appear in genomics, in biological characters, and in the fossil record.*
De Novo Genes: One prediction of evolution is that de novo genes should have predecessors via a mutational pathway. This prediction drove the research that shows how the “anti-freeze” gene in Arctic codfishes evolved 13-18Mya from non-coding DNA. Source
Genomic distribution of mutation types. As @glipsnort demonstrates, evolution predicts that the frequency of mutation types can be extrapolated across hundreds of thousands of generations once populations diverge. This distribution of mutation frequencies is observed empirically.
Biological systems: The 2009 Nobel Prize in Chemistry was awarded to research showing that ribosomes are ribozymes. This was a prediction rooted in evolutionary theory. Source
Pseudo-genes. If dietary sources of vitamin C are available in abundance, the negative selection against vitamin C gene breaking changes should disappear. The broken vitamin C pseudo-gene has in fact been observed in most primates.
Evo-devo. Evolution predicts that many phenotypic changes across time are the result of changes in regulatory gene networks. Thus we would expect much greater conformance in structures at zygote and early fetal stages than at later stages of development (after regulatory gene networks have done their work). This is indeed observed empirically.
Sub-optimal engineering designs. Moreover, biologists expect evolution-driven development process to sometimes result in structures which are sub-optimal from the engineering perspective as long as they result in a net advantage to a population. This is observed in the laryngeal nerve of the giraffe, which extends meters beyond the shortest path (and incurs a cost in reduced vocalization ability). Source
I am sure that a biologist like @DennisVenema could provide other examples of evolution predictions, if need be. I’m just a data scientist who likes to discuss biology.
* Why is there a relational structure at all between species like whales and hippos, or even between chickens and primates (I’m thinking of the vitellogenin pseudo-gene)? Inheritance and divergence are observed in biological systems today, so they provide the parsimonious explanation. On the other hand, a dependency graph model introduces scores of extra parameters, which in turn permits overfitting. Due to this ability to overfit, dependency graphs can “fit” non-sequential genomic data quite well, but they spectacularly fail the parsimony criterion for scientific research. In addition, dependency graphs are for the foreseeable future incapable of making predictions for many classes of biological data; hence they are for the foreseeable future incapable of serving as the basis for a theory of biology.