"Open up the algorithm" is a fantasy where you point a function at a tweet and it returns a boolean. In reality, content moderation at this scale is a complex problem that requires a sophisticated system of many parallel strategies (some of which are more automatable, some less) and many thousands of people to manage.

Is in your opinion the Twitter moderation algorithm more complicated than the Linux Kernel or Bitcoin software code bases? Those are audited and modified publicly al the time. The human part can also be audited if the rules are stated clearly and an explanation is always given when an action is taken.