A useful way to break things down is by forms of reasoning:

- **Deductive reasoning:** infer the logical implications of some premises or statements.
- **Inductive reasoning:** make a generalisation or future predictions from some premises or evidence.
- **Abductive reasoning:** try to figure out the best explanation for the evidence - typically the explanation with the greatest [explanatory power](https://en.wikipedia.org/wiki/Explanatory_power).

Maths and logic use deductive reasoning.

Determining what someone would do in some situation could be based on inductive or abductive reasoning.

Theories of gravity, and science in general, is inductive or abductive, not deductive.

---

Alongside and related to this, foundational parts of producing accurate predictions are:

- **Empiricism** (the scientific method, broadly speaking)
  - You use observations to get a baseline understanding of reality.
  - You take your reasoning and test it in the observable world, to verify that it actually works, and to fix or improve it.
- **Consistency and non-contradiction**
  - For reasoning to be sound, it should produce the correct results when applied consistently across a number of different questions. If reasoning is uniquely used to try to draw one conclusion, and it applies nowhere else, it's functionally useless and entirely untrustworthy.
  - Reasoning that produces contradictions is not sound.

If you determine what someone else would do based solely on what you'd do, you're often going to be wrong, as different people have different motivations and preferences and ways of thinking, driven by their biology and their life experiences. By refining this through empiricism, one can predict this much more accurately, by considering what you know about differences in how people think, and what you know about the specific person in question, by considering your motivations compared with theirs, etc.