Argumentation schemes should be generated by a mixture of top-down theorizing and bottom-up abstraction, should be both descriptive and prescriptive, can all be regimented into species of a single one-premiss super-scheme, should be as general as is consistent with being readily applicable, have three kinds of associated critical questions (concerning their premiss, their assumptions, and exceptions), are the same as reasoning schemes, and in general straddle the distinction between conclusive and non-conclusive support. Appeals to considerations or criteria, sometimes called ‘conductive reasoning’ or ‘conductive arguments’, are better treated in a dynamic way that takes account of the subjectivity of their authors and critics, rather than with the static approach of argumentation schemes and critical questions. One can reasonably take into account diverse considerations and disagreements about them by comprehensively listing considerations, assessing their acceptability and relevance, reframing, adjusting the option space, debiasing, estimating importance, and allocating the burden of proof. Reasoning by analogy projects a queried property from one or more source cases to a target case on the basis of one or more assumed similarities. It can be validated either by a determination relation or by recognition that the source cases have the queried property in virtue of having the assumed similarities or (weakly) by the number and variety of shared similarities and small number of dissimilarities. Practical reasoning, i.e. reasoning about what is to be done, takes many forms: not just means-end reasoning, but also application of a general principle to a particular case, adoption of a policy on the basis of relevant considerations, appeal to consequences, and so on. John Pollock’s model of practical reasoning is superior to others in taking into account likings and dislikings as well as beliefs, desires and intentions. But it is deficient in making no allowance for communication between rational agents, social cooperation, or moral constraints. Formal systems for deliberation dialogue (i.e. dialogue about what is to be done) should allow for a variety of types of non-factual input, not just goals but also preferences, values and constraints. Likewise decision-support systems should allow for a variety of argumentation schemes and should be transparent to their users about the argumentation used to produce a recommendation. Means-end reasoning involves not only selecting a goal and discovering a means that would achieve it but also ascertaining that the goal is achievable, that the means is permissible, that no alternative means is preferable, and that the side effects do not outweigh the benefits of achieving the goal.
