My personal view is that calling models or agents personas or users really misses the distinction between human users and synthetic models and agents. I have suggested to some of my colleagues that we need to understand the flipside of traditional human factors is that we now need to include requirements, guidelines, and techniques to determine what these synthetic need to receive as inputs to optimiW their support to human users of our systems and organizations. Most of these requirements are for things that we take for granted in humans, e.g. when background noise is high and you and I have a conversation, i may mean in to hear what you are saying, but will a synthetic entity realize that? What does that entity need as inputs to realize that it should increase its volume? Or, when I use speech recognition as an input modality, what are the performance requirements ro ensure that a synthetic agent can provide adequate support usong that modality?
My view is that they are neither conventional systems nor traditional users. They occupy a distinct middle position, functioning as intermediaries that act with limited autonomy. In a typical system or algorithm, the human remains the primary decision-maker, whether through direct input or predefined logic. It is still an “if this, then that” model. AI operating as a user introduces a different dynamic because it performs its own assessments and selects actions within defined boundaries. This is why modeling and scoping their behavior is essential; it captures what these non-human actors contribute as quasi-users within a larger information ecosystem.
Love this. We finally admit there are two users now: humans AND the models.
IXA is basically “UX for machines”, and honestly… it’s overdue 😅
I actually presented at HFES Healthcare Symposium last year about the "non-human personas." This is an extension of this idea.
My personal view is that calling models or agents personas or users really misses the distinction between human users and synthetic models and agents. I have suggested to some of my colleagues that we need to understand the flipside of traditional human factors is that we now need to include requirements, guidelines, and techniques to determine what these synthetic need to receive as inputs to optimiW their support to human users of our systems and organizations. Most of these requirements are for things that we take for granted in humans, e.g. when background noise is high and you and I have a conversation, i may mean in to hear what you are saying, but will a synthetic entity realize that? What does that entity need as inputs to realize that it should increase its volume? Or, when I use speech recognition as an input modality, what are the performance requirements ro ensure that a synthetic agent can provide adequate support usong that modality?
My view is that they are neither conventional systems nor traditional users. They occupy a distinct middle position, functioning as intermediaries that act with limited autonomy. In a typical system or algorithm, the human remains the primary decision-maker, whether through direct input or predefined logic. It is still an “if this, then that” model. AI operating as a user introduces a different dynamic because it performs its own assessments and selects actions within defined boundaries. This is why modeling and scoping their behavior is essential; it captures what these non-human actors contribute as quasi-users within a larger information ecosystem.