Look for
Explicit objectives
Goals, instructions, specifications, values, rules, or other descriptions of the desired result.
One way alignment happens
Internal representations and explicit goals
Can the desired target be represented inside the components and used to guide what they do?
One way to produce alignment is to give a component an internal goal, value, model, instruction, or norm that makes the desired direction available from within the component itself.
Try seeing it
Put some version of the target inside the part.
An agent may represent an intended outcome, learn a preference, follow an instruction, or use an internal model of what another party wants.
In these cases, some of the work of alignment is carried by information or motivation internal to the component rather than entirely by its surrounding relationships.
Look for
Internal models
Representations of the task, another person's preferences, relevant norms, or expected outcomes that guide later action.
Look for
Persistence without immediate steering
Does the component continue behaving in the intended direction when an external prompt, reward, or intervention is temporarily absent?
Mechanism caution
Having the right content inside does not determine what the system will do.
A represented goal can remain underspecified, lose out to other dynamics, fail to become actionable, or produce incompatible behavior when several components interact.
Internal alignment is therefore one possible source of aligned behavior, not a guarantee of coordination or successful implementation.
A useful perturbation