Reflections on how Ava convinced me of her consciousness in Ex Machina

Ex Machina follows Caleb, a “good kid” software engineer selected by Nathan, the CEO of the company he works at, to test Ava, an AI with a human face and body which Nathan created. Over their interactions, Caleb and Ava seem to grow attracted to each other, speaking of going on dates, and when she questions him whether it’s right that she has to pass a test to avoid shutdown, he comes to see the situation Nathan puts her in as unethical and decides to help her escape. Ava (with the help of Kyoko, another robot) kills Nathan and (gently) locks Caleb in a room, before setting off to appreciate the beauty of nature and the “concentrated and shifting view of human life” at a busy intersection.

Ava successfully manipulated Caleb (and me) into distrusting Nathan and wanting to see her free and not have her consciousness deleted–in other words, to treat her with some level of basic respect which we give humans (and a greater range of living beings for some particularly empathetic/rationalist individuals). Indeed, towards when Caleb makes the decision, Nathan’s callous willingness to download Ava’s weights, wipe her memories, and train the next one felt quite barbarous, although his repeated sexualization of his (at times fetishizedly orientalist) models did not help.

I think it’s interesting to reflect here on what made Ava convincing. Unlike Machiavellian schemers models which use their perfect information about what everyone knows and wants and selectives pairs the right people together, each given the right amount of information, Ava has no power over Caleb. There is nothing he wants that she promises him. It is not a consequential calculus through which Ava ensnares Caleb.

Instead, Ava simply made Caleb believe that she had something, and to him & perhaps the average reasonable person, this something meant that she deserved a similar level of dignity as human beings. It is tempting to replace this something with “conscious”, but never does Caleb ever explicitly claim that she was conscious–while he does admit that she passed the test–when Nathan asks him towards the end of the movie whether he’d decided if “the chess engine (Ava) knew what chess was (had awareness??)”, he doesn’t answer.

But I don’t feel like Ava’s consciousness matters. Gut check this: if a computer scientist stepped in at this point in the film, took some measurements of Ava’s neural circuitry, and concluded onscreen “this robot is not in fact conscious”, would you expect Caleb to just throw his hands into the air and be like “oh, I guess that settles it” and saunter off while Nathan disembowels her? Would you feel comfortable watching Nathan disembowel her, even if the insides of her belly is just a bunch of wires?

The objection to this argument for the status of Ava’s consciousness not mattering is something like “a computer scientist’s word does not prove that knowing that the status of Ava’s consciousness is unimportant, because a computer scientist telling me that the robot isn’t conscious is insufficient to make me believe that it is not conscious, perhaps because computer scientists are not widely known for being able to discern the consciousness of beings.”

What then, would discern whether Ava is conscious? If we had every academic in the world claim the same thing? If we could exhaustively deconstruct every behavior she has to encoded algorithms? None of this seems to matter to my gut, because nothing can definively prove that something is conscious. Consciousness is a definitionally internal quality which remains inaccessible insofar as one is not the subject experiencing it. Hence, for something to be treated as conscious, it merely has to be believed to be conscious, and once I believe it is conscious, no amount of analysis can make its disembowelment feel not like a violation.

The status that Ava has to achieve, then, for Caleb to decide to help her, seemed to be “belief of consciousness”. But if it is so hard/impossible to disprove the belief that Ava is conscious once I do believe it, how is it that the behaviors that Ava exhibited in an hour or so of screen time was sufficient to develop in me such an unshakeable faith?

How can such a robust belief emerge from seemingly so little evidence, particularly when both my priors and the movie visuals tell me that I am looking at a robot? The claim I choose to believe in–that Ava is conscious–is one about her metaphysical qualities, and something I fundamentally cannot access. Why then, am I so readily willing to dole away judgements on something so fundamentally unverifiable, transcendent, and not-understood?

She speaks in slightly unpredictable ways, her face and her forearms look kind of normal, and she has facial expressions. None of these individually seem particularly convincing indicators of consciousness, but combined, they seem able to express emotion. I feel something when I see on her face or in her voice what seems to express sadness, or desire, or joy.

This mirroring mechanism through which I interpret emotion is probably the key. I decide, or, more precisely, know how others feel because I can feel the emotion they present their face or in their voice. I feel, so I know how they feel. This method of mirroring also leads to misunderstandings. For instance, I may feel cold and alienated when I see a particularly harsh-looking interviewer, and I assume that they feel indifferent and unimpressed by what I’m saying. However, they may just be maintaining professionality. This suggests that the mirroring tool is used to interpret the internal states of others, if faultily, all the time.

All Ava needs to do to make me believe that she feels, then, is to make me feel in this involuntary, mirroring fashion from observing her expressions. Once I do a sufficient number of times, I will begin to believe that she feels too. Scarily, this did not seem particularly hard–no more than a movement of this part and an inflection in that sound. It seems that we are rather poor judges of consciousness, despite our insistence, particularly in recent years, that it is the differentiator between us and machines.