3

The recent Google Duplex demo caused a firestorm of criticism online.

One of the things that I kept hearing was that it was an unethical deception for an AI system to present itself as a human caller.

I’ve given it some thought and I can’t figure out just what is immoral or unethical about this.

Some people have even argued that Google set out to deceive intentionally, but I fail to understand how that could be, insofar as the whole point of the system is to make the experience seamless for the callee, so that the callee’s experience of the interaction is as easy as if it were a human on the other end.

Also I don’t understand where the idea that the callee has some kind of moral right to know if they are interacting with a computer system instead of a person comes from. Can someone elucidate this? Why in this interaction does that right exist, compared with for example the right to know other facts about a caller if that caller were human?

Legally and morally, businesses have an obligation to serve customers without prejudice. For example it would be wrong to refuse a customer service because of their race or religion.

In terms of the interaction, I cannot think of a reason to frame as unethical the sophistication of the AI unless the knowledge of the non-human identity of the caller would cause the callee to behave differently upon that realization.

And then, for me, the question becomes what that difference in behavior entails. The only thing I can think of is simply to hang up on the caller, the desire to be able to refuse to interact with an AI.

Can someone explain to me the counter argument? Why is it unethical in this case that a business would not know that a caller is an AI and not a human?

4
  • This seems mainly opinion-based, however, I hope someone can provide an answer that is not just giving their own opinion. May 13, 2018 at 22:17
  • @FrankHubeny mainstream media like the BBC and others are quoting professors in philosophy as saying that this is deception and unethical. Unless you’re suggesting that these people are charlatans, then there must be strong arguments for their position, no? May 13, 2018 at 22:23
  • I voted to close as primarily opinion based. May 13, 2018 at 23:42
  • @FrankHubeny I strongly disagree that this is an opinion based question. I think this is a relevant and up-to-date question for modern philosophy. May 14, 2018 at 0:18

3 Answers 3

2

"Thou shalt not make a machine in the likeness of a human mind" - Orange Catholic Bible

The term "unethical" is often thrown around in non-philosophical venues to mean a variety of undesirable things. Often they have very little to do with ethics, so one has to actually ask what people mean when they claim it is unethical.

One aspect I could see as unethical is that it is typically considered unethical to lie. If someone believes they are speaking to a human when they are actually speaking to an AI, there's a bit of a lie by omission when the AI fails to identify itself. However, there's a long series of debates as to whether lies by omissions are the same as activly lying.

Another potential issue is that it is generally assumed that conversations are not recorded unless this is specified ("This call may be recorded for quality assurance."). It is yet unclear how AI language processing fits into this.

I would not consider this protected in the way race or religion is. Those are protected traits that are treated as a very different class from every other distinguishing factor a customer may have.

4
  • I’m not sure the lie by omission argument is persuasive. It seems strange to me to consider what facts about the caller are relevant from a moral point of view. Normally when a human calls, we don’t expect the callee to know the gender of the person, or what they are wearing, or even if they are calling on behalf of someone. Often my wife instructs me to make calls where the call is for her benefit but I don’t tell the business on the other end because it’s clearly not relevant. Same with when I call a toy store to find out its hours because I’m buying a gift for a child. May 13, 2018 at 21:46
  • I didn't say they were persuasive arguments, but they are arguments that are at least reasonable (in my mind). There are indeed codes and expectations when we pick up the phone. Google Duplex absolutely violates some of them. Whether that makes them unethical or not is another question, but there's no doubt in my mind that it violates social rules. Most of our societies rules are not designed to deal with AIs that can pass the Turing test.
    – Cort Ammon
    May 13, 2018 at 21:49
  • The recording aspect is more interesting but I disagree that businesses should have an expectation that a phone call on a business number would be private. This is not a call between friends or individuals. This is a public business where the expectation of privacy seems less to me. As to protected classes, if the person using Duplex has a physical disability which makes it impossible for them to make a call on their own without assistance, refusing to interact with Duplex seems like a real violation of rights. May 13, 2018 at 21:51
  • which social rules does it violate? May 13, 2018 at 21:54
1

un-ethical - The ethics appear to be not declaring something that might cause offence or be upsetting to the other party, for whatever reason, without first allowing the other party the chance to decline such interaction.

I-robot film demonstrated such a bias, against the process of interaction with AI.

In reality talking to a computer or a person who is doing a simple task of obtaining information and giving a need, is not an ethical issue by information exchange where no threat or problem will arise from the communication.

If on the other hand you were talking to someone who had done great harm to you or someone you knew, and if you knew this was the person, you would not talk, then that would be called unethical, because the response would be upsetting, and the denial of contact would be 100%.

Some people are obviously scared of this means of imitating human communication though it is irrelevant. Some hate the press one for x, press to for y, which is a human voice giving you a list of options, which is similar. But equally some people hate this and want a more complex solution.

So to answer the question, if someone hates this type of interaction, it would be unethical not to declare the type of interaction people are participating in so they can choose to stop or carry on. This is similar to recording phone calls preceded by a warning this is happening. It is not a new ethical dilemma with a simple solution.

7
  • Ok exactly. The last paragraph says it all. If what you’re saying is true, then how does this fly with people who don’t want to talk to homosexuals or people of color or disabled people? Specifically, in the case of a business, it should not have the option of refusing service to someone merely because of how they communicate with said business. If the business has a published phone number, then they shouldn’t get to hang up on a legitimate call merely because the call is made by an AI. These aren’t crank calls, and this isn’t a private interaction. May 13, 2018 at 22:36
  • 1
    Said another way: the decision to take into account the non-human nature of the call and thereby refuse service can’t possibly be sole grounds for claiming that it is deceptive for the AI not to identify itself as such, otherwise it would be similarly deceptive for a person of color to call a white supremacist’s business and not identify themselves as a PoC on the call merely to give the business owner a chance to slam the phone down. May 13, 2018 at 22:40
  • This is simply a pr issue. Why upset people, if they do not want to be involved. Over time they will get used to it. A company is not employing people but using AI because it is cheaper. And the customer is always right, as they are the purchaser or provider of an asset. If on the other hand it is a business gaining custom, it would be illogical to turn AI away. I suggest the discrimination argument is not a valid parallel. AI have no civil rights being infringed or hurt feelings!!!!
    – PeterJens
    May 13, 2018 at 22:50
  • regarding the last point: a human will has asked the AI to carry out a task for it. It’s not just a computer acting randomly. Discriminating against it is arguably discrimination against the person who tasked it. May 13, 2018 at 23:06
  • @ribaldEddie Are you saying AI is a victim of discrimination which should be defended in the courts? Only humans can be discriminated against. A person is not talking to the programmer of a system, but the system itself. And you cannot discriminate against a manager. You are talking way outside discrimination law. And what rights are being infringed for the AI? Name one.
    – PeterJens
    May 13, 2018 at 23:13
0

At a minimum, I believe Google's decision-makers know that humans place a higher value on interaction with other humans vs. AI. So at a minimum, they would be knowingly benefiting from deception, in the form of receiving (or selling the opportunity to receive) fraudulent goodwill.

Perhaps the biggest harm would come from how this fraud would impact third parties... Some people like to have a human on the other end of the phone, and some companies currently hire a lot of human beings to answer phones in order to earn that corporate goodwill from their customers.

It would not necessarily be unethical to displace these workers with a better or more efficient product or service. But to displace them with a false service that gives the false impression of talking to a human being is not the same as that. This would be more equivalent to selling a physical product to the people of Chicago, under the advertisement "Buy direct from our Chicago factory!" When it turns out the product is actually just shipped to that Chicago factory for sale, rather than made there as intentionally implied.

EDIT: The facts of the first sentence regarding human perceived value of interaction with other humans vs. AI was disputed. Brief online research showed many articles concurring, and none that I could find disputing the value statement of the first sentence. A representative example of available data is here: https://www.prnewswire.com/news-releases/nearly-90-of-people-prefer-speaking-to-a-live-customer-service-agent-on-the-phone-despite-efficiency-of-phone-menus-for-businesses-300830525.html

5
  • Your first paragraph relies on a non-sequitur. Humans place value on human interaction sure, but that doesn’t mean that they don’t place value on non-human interaction. The idea that there’s deception to obtain goodwill would only work if you specifically mean to say that humans place no or less value on interacting with Duplex. But that’s not what you said. I see no reason to assume that humans place no or less value on interacting with AI— AI has not been around for long enough for most people to know what they think. May 21, 2022 at 6:26
  • I don't think what you describe is a non-sequitur, as much as a simple dispute of the facts. (Which is important here, because incorrect facts and non-sequiturs require different handling.) Anyway, the thing you find incorrect... The idea that people prefer to speak to humans over AI, is testable. And polls have been taken / tests have been done. For one of many examples, see: prnewswire.com/news-releases/… May 22, 2022 at 16:57
  • we are not talking about automated prompts. We are talking about having a conversation / interaction with a system that is indistinguishable from a person. Given the glaring issue with your claim, I don’t think you’re posting in good faith. May 22, 2022 at 17:03
  • I clarified the first sentence too. Good point on including both sides in the statement. On the dispute, the fact is that if you ask people if they prefer speaking to another human or AI, they (by a wide margin) say they prefer speaking to other humans. So, if you secretly replace those humans you are committing fraud. It doesn't matter if the AI system can perfectly mimic a human or even if it is much better than a human for the job at hand. It is not a human and pretending it is, is fraud. Fraud is unethical even if intent is good, as it violates a person's self-agency. May 22, 2022 at 17:16
  • again, you haven’t linked to a survey that shows what you claim. Your survey link is like asking people if they enjoy a cross-country drive that takes a week and a half and then extrapolating that to assume they hold the same enjoyment for a flight that takes four hours. Sure, both will get you to the other side of the country but they aren’t the same trip. As to the claim of “fraud”, fraud is a legal term. I don’t see that applying here. The underlying premise is that business owners have a right to know with whom they are speaking on the phone but on its face that isn’t true. May 22, 2022 at 17:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .