Artificial agents’ explainability to support trust: considerations on timing and context