Scheme is a functional programming language and one of the three major dialects of the programming language Lisp. (The other two being Common Lisp and Clojure.)
Pro Simple syntax
Scheme syntax is extremely regular and easy to pick up. A formal specification of the syntax fits onto just a few pages; it can be introduced informally in a paragraph or two. Users are not distracted by remembering how to write if statements or loops or even operator precedence because every syntactic follows the same pattern.
Ultimately, everything looks something like this:
(func a b c)
This includes not only user-defined functions but even control flow:
(if cond then-clause else-clause)
or even primitive operations like
(define foo 10) (set! foo 11)
This means that nothing really has special syntactic treatment in the language. There are essentially no weird edge-cases to memorize, and different concepts are given a more equal weight in the language. (Unlike Algol-like languages which tend to given undue weight to loops and assignment statements, for example.)
Unlike most languages, Scheme actually accords both functional programming and imperative programming roughly equal status. Many other languages like Python and Java are staunchly imperative while SML and Haskell are primarily functional; Scheme is a nice middle ground.
Additionally, since Scheme syntax is extremely flexible, it can easily be re-purposed for teaching non-deterministic and logic programming. There is no need to learn a new language like Prolog when the same ideas can easily be expressed with Scheme syntax.
Pro Robust metaprogramming
The quotation functionality of Lisp allow for extremely powerful, yet syntactically straightforward metaprogramming via macros. This is more powerful than the C preprocessor while being less involved than something like Template Haskell or F# quotations.
Using macros to properly decompose a problem domain teaches new developers good habits, improving composibility and reliability when tackling large programs. Scheme metaprogramming also serves as a gentle introduction to domain specific languages.
Pro Great, well known textbooks
There is a set of very strong textbooks introducing CS and programming using Scheme. These books are available for free online.
The most famous example--and one of the most famous CS books full stop--is Structure and Interpretation of Computer Programs usually known as SICP. This book introduces fundamental ideas in computer science and covers an incredible amount of material quickly and clearly without requiring any prior knowledge.
However, some people find SICP a bit challenging as a first text. Happily, there are other more introductory texts as well. Simply Scheme is a book designed to be read before SICP, easing beginners into the language as well as CS and programming in general. How to Design Programs is another text used in introductory college courses.
Pro No Magic - it's clear how everything works
Scheme has far less built into the language itself, helping students see that things like OOP are not magical: they are just patterns for organizing code. Everything in Scheme is built up from a very small set of primitives which compose in a natural and intuitive fashion.
Having a language that does not accord many things special status helps keep students open minded. This will help students later go between different languages and paradigms from procedural to object-oriented to functional.
Con Very different semantics from mainstream programming languages
Con Fragmented ecosystem
Scheme is an IEEE standard, not an implementation. Unfortunately, the standard is too minimal and practical implementations have diverged--they had to expand on the standard to get anything done, but did so in incompatible ways.
The later de facto standard R6RS tried to correct this, but lost Scheme's minimalist elegance in the process. The newer R7RS standard takes the best of both worlds with an elegant minimalist core and a practical standard library.