Interviews can describe a working relationship without creating one. A trial project lets people encounter a small part of the real thing: clarifying a task, producing work, responding to feedback, and deciding how to proceed when something is unclear.
Automattic uses paid, part-time trial work in hiring. The important distinction is that a trial is bounded work with explicit terms, not an indefinite audition in which the participant keeps proving themselves without knowing when a decision will arrive.
Both sides are learning
Imagine a small organization hiring someone to improve its volunteer handbook. A limited paid project could reveal how the applicant handles incomplete information, and show the applicant whether the organization's feedback is useful or chaotic. The trial generates evidence about the relationship, not only about one person's output.
The arrangement still favors people able to make time. A candidate with an existing job or caring responsibilities may find even a paid part-time project difficult to schedule. A badly scoped trial can also become production work obtained from many applicants while only one receives an ongoing opportunity.
Scope, payment, access, and decision timing therefore shape the quality of the experiment. A trial should give both parties a real but limited experience of collaboration, with an understandable ending even when they decide not to continue.