Autor: Nermin Sefić
A model can prepare a draft, but cannot take responsibility for a decision's consequence. A commentary by Nermin Sefić.
A model can prepare a draft, but it cannot take responsibility for the consequence of a decision.
AI can find a pattern, summarise a document, and suggest text. That's useful, but it's not the same as making a decision. The final decision must remain with a person who understands the context, holds the authority, and can explain why a given result was accepted or rejected. An organisation that hides accountability behind a model isn't automating a process. It's automating an excuse.
This publication is an original analysis; the sources cited serve as reference documentation.
In practice, this means the accountable person must be able to reject a model's recommendation without justification, because that very ability to refuse is the proof that the decision genuinely belongs to a person, not the algorithm that proposed it.
The distinction matters most in edge cases, where a model's training data may not reflect a genuinely novel situation, and where a purely mechanical application of past patterns could produce a confidently wrong result. A human reviewer who understands this limitation, and who is empowered to override the model without needing to justify every override, is not a redundant layer of process — it is the mechanism by which the organisation catches the cases the model was never equipped to handle correctly.
#AIAccountability #HumanOversight #BusinessDecisions #NerminSefic #GNKASG #GNKDINAMOLtd
Cjelovit tekst i izvor: https://gnk-asg.hr/en/commentary/ai-must-not-write-the-final-decision/
Autor i urednička odgovornost: Nermin Sefić. Izdavač: GNK ASG d.o.o..
#GNKASG #GNKDINAMOLtd #NerminSefic #BusinessIntelligence #NerminSefić #GNKASGdoo #Zagreb
