diff options
Diffstat (limited to 'paper/background.tex')
| -rw-r--r-- | paper/background.tex | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/paper/background.tex b/paper/background.tex index 9457b35..08b30ca 100644 --- a/paper/background.tex +++ b/paper/background.tex @@ -7,7 +7,7 @@ Several programming tutors generate hints from differences between the student†Rivers and Koedinger compare students’ programs directly to a database of previous correct submissions~\cite{rivers2015data-driven}. They reduce program variability using equivalence-preserving transformations, such as inlining functions and reordering binary expressions. Hints are generated by suggesting a minimal correct step leading from the current submission to the closest correct program. % unit tests -Another option is to compare program behavior. Nguyen et al. classify programs according to results on a preselected set of test inputs~\cite{nguyen2014codewebs}. Li et al. generate test cases to distinguish between programs by selecting inputs that exercise different code paths in the program~\cite{li2016measuring}. Such tutors can point out pertinent failing test cases, which can be very helpful. However, to do any sort of conceptual analysis of student programs, we need to define some language for describing invariant properties of programs. +Another option is to compare program behavior. Nguyen et al. classify programming mistakes according to results on a preselected set of test inputs~\cite{nguyen2014codewebs}. Li et al. generate test cases to distinguish between programs by selecting inputs that exercise different code paths in the program~\cite{li2016measuring}. Such tutors can point out pertinent failing test cases, which can be very helpful. % constraints \emph{Constraints}~\cite{mitrovic2012fifteen} encode domain principles using if-then rules with relevance and satisfaction conditions, e.g. “if a function has a non-void return type, then it must have a return statement”~\cite{holland2009j-latte}. If a program violates a constraint, the tutor displays a predefined message. Le’s Prolog tutor improves constraint-based diagnosis by assigning weights to different types of constraints~\cite{le2009using}. |