WebGreeter Fails Turing Test

Beware, if you visit WebGreeter.com a disturbing thing will happen. You will be immediately accosted by what appears to be a chatbot, but is apparently a human doing a creepy impression of a chatbot. This cyborg thing will ask you for your contact information. If you give it to them, they may use it right away to call you on the phone.

Live Attractive Smiling Operator! Click Now!

Live Attractive Smiling Operator! Click Now!

Based on their website, and the high pressure salesman who called me without an invitation, they believe in an aggressive approach to sales. The salesman even used the word “aggressive” several times in his pitch.

The first time I encountered this technology was on a website for expert witnesses. I had trouble using the site and clicked on the “Live Operator” icon. A chat window opened, but everything that they said to me appeared to have come from a script, while seeming to ignore the gist of my questions. I was kind of offended, since this cyborg thing insisted it was a live operator (each time using exactly the same perfectly worded sentence to convey that) while continuing to answer a different question than the one that I asked. At one point I challenged it to type something instead of pasting a canned response, to which I got an obviously canned response that it was not allowed to engage in personal conversation.

(What? Isn’t that the point of a live operator? The ability to engage on a personal level, human-to-human, empathically? Do they think I want a live operator so that I can be ignored in person?)

I concluded that it was not human. But when I went to the WebGreeter website to learn about this weird chatbot technology, I found out it really HAD been a human. That human failed the Turing Test, by successfully declining to use human communication skills with me.

Folks, this is what I’m complaining about with scripting, scripted testing, scripted behavior of any kind. My situation required an appropriately trained and reasonably motivated human to listen to my questions and help me solve my problem. Instead, I meet someone’s idea of an efficient technology that creepily removed what’s good about humanity while being able, technically, to claim humanity. The human in that equation could hide behind rules, pretending to help me while rendering no assistance at all.

I don’t mind that they have scripts. I mind when they are followed instead of applied. This distinction is crucial. In applying, the human remains in charge. In following, the human actually lobotomizes himself in an effort to become animated furniture.

When we deal with people, we ought to be able to trust in certain fundamental human qualities. But just as this crazy WebGreeter company has thrown those out the window, so too have many “process” people torn them out of software projects. Or tried to. Instead they want us to perform rituals. Nice repeatable rituals.

40 Responses to “WebGreeter Fails Turing Test”

  1. Zach Says:

    Hi Guys,

    Nice post James. Please continue to visit webgreeter.com to chat with unique chatbots :)
    I wish you luck trying to prove we are unreal. We appreciate your business.

    [James' Reply: See, this is what I'm talking about. I haven't done any business with Webgreeter, nor am I likely to. What I'm doing is ridiculing them. This is an inappropriate response-- almost as if the author were following a poorly designed script.]

  2. Oliver Smith Says:

    Interesting Alan Turing didn’t actually state that the now popular “Turing Test” was a test for inteligence in his original paper. Infact I believe that it’s been a big failure to the field of AI.

    [James' Reply: It was, for him, a test of whether a machine can think. But in his test he assumes a basic competence on the part of the human that most humans don't, in fact, possess. Humans can think and most of them think quite well. But most humans are unmotivated and untrained as testers and that's where the Turing Test falls down.

    If one human convinces other humans that spaceships are about to come and take away the souls of true believers, therefore we should commit suicide tonight at 6pm. I don't think "wow, that was a persuasive argument." Instead I think "those were really weak minded humans." Similarly, when I hear that a computer has passed the Turing Test, I think the testers weren't very good.]

  3. Joe Strazzere Says:

    Great analogy, James! “Don’t think – just follow the script! Don’t think – just test!”

    And I’ll bet these automatons at WebGreeter are maximizing every one of the metrics their managers have most likely imposed on them.

    (If I recall correctly, the Turing Test is designed to test a machine, not a human. So technically, a human cannot fail the Turing Test, right?)

    -joe

    [James' Reply: The Turing Test was designed to compare a human to a machine, actually, on the presumption that a human can think. It's perfectly legitimate for a human to be subjected to the same test. But when that happens, it challenges our notion of what human means.]

  4. Paul A Houle Says:

    It’s hard to be scripted and spontaneous at the same time. That’s why people like Brad Pitt get paid big money!

  5. Kevin Says:

    In business, there are business processes. These processes are the rituals that smart people have discovered to be easy and repeatable.

    [James' Reply: I don't agree with this as a general statement. This trivializes the whole idea of process. What evidence have we that "smart people" came up with any given process, or that they discovered it to be "easy and repeatable?"]

    The fact that they are easy and repeatable and effective make them usable by any idiot. Idiots are cheaper than smart people, hence the reason that you build business processes.

    [James' Reply: Any idiot? Any? How could you make such a claim? What evidence can you offer?]

    The fact that McDonalds can make the same hamburger in a predictable and efficient way comes from a business process.

    [James' Reply: First, we don't know that McDonald's can make the same hamburger, nor that it does so in an efficient way. Second, even if they do, we don't know that it can do that because of their "business processes." Maybe it does it because of the inherent humanity of their staff.]

    This franchising process has been the reason that many businesses have succeeded. If you were interested in business and the success of the business that you are creating, you yourself would be planning a way to make business processes in order to insure your success. You would also not have had time to make this unintelligent article.

    [James' Reply: Congratulations, you have faithfully regurgitated the myth of business processes, just as you were programmed to do. But where is your evidence? Nowhere! How do you respond to the points I raised? You don't! Don't speak to me in the guise of a robot, speak to me as a man.

    I suppose you missed it, but the whole point of my post was to point out that the scripts these WebGreeter people were following did not work. The problem of responding to humans in a way that makes them feel heard is a different problem from reproducing a food product, don't you think?]

  6. wgee Says:

    I would still say that this is technically not a human, since it is scripted and following rules it is a system. See John Searle’s “Chinese Room” for more info

  7. Thomas V Says:

    Are you familiar with Searle’s Chinese room experiment? (It falls more in the realm of metaphysics than epistemology.)

    Basically, the thought experiment is to imagine a machine that seems to understand Chinese – but is in fact implemented by a roomful of people who have been given detailed instructions, written in English, about how to manipulate Chinese symbols to appear to have a conversation – a very sophisticated AI implemented in a low-tech way. No one in the room understands Chinese. Can we really say that the room understands Chinese?

    In this case, the same mechanism has been used but the system has been implemented in such a way as it takes a human being with a mind and effectively obscures that fact. I wonder if proponents of the position “Hard AI” would say therefore the system really lacks intelligence despite being designed devised by and consisting of actual intelligences?

    [James' Reply: Yes that's a famous philosophical puzzle. It relates to the mind/body duality problem (is the mind a projection of the body, or is it something else?) and generally raises the question of what can be automated and the meaning of automation.

    Like a lot of philosophy, I approach it as a testing problem. (I think that means I'm a pragmatist in the original sense of that word.)

    If such a room could be built, then the question "does the room understand Chinese?" could only be settled as a matter of testing. But to do so, the tester would need to know something about how it works, and how it might fail. Otherwise, it will be a rather shallow test. For instance, there may be certain Chinese ideograms that the human inside the room routinely miscategorized in a way that would dramatically altered the meaning of a sentence in a manner that would be extremely unlikely for a native speaker of Chinese. To test for that, we would want to cover all the ideograms.

    Our confidence in the Chinese room is therefore dramatically-- and quite rationally-- altered once we are clued into the possibility that this system is not human (since humans are entities whose failure patterns we believe we understand and accept).]

  8. Mr Flibberly Says:

    How do you feel about that?

    [James' Reply: Thank you, Eliza.]

  9. j4m3z Says:

    I worked at a call center years ago where we dabbled in live Web chat and email customer support in its infancy. We were not allowed to use any words that weren’t on the script – either live or in email responses. It seemed a very strange policy really, but apparently is still in practice. I know my own bank’s Web chat has a lot more lenient conversation guidelines and I certainly prefer it that way.

  10. Steve Campbell Says:

    It would be a funny story, except that it is so frustrating when it happens. I cannot recall which company it was, but I had a similar experience. I also asked the question “are you a real person”, to which the response was somewhat “annoyed” in tone, thus allowing me to conclude (or at least have a working theory) that it was a person.

    I think it may be more directly advantageous to the bottom-line to advertise “live support” than it is to actually implement it.

    [James' Reply: Somehow this reminds me of the time I ordered a sandwich in Amsterdam. I wanted a Reuben, but in his little shop I didn't see any means of cooking, so I asked the guy if his Reubens were "real," as in hot. He snorted, and said "Yes sir, all of our food is real!" He then wasted no opportunity to make a stinging sarcastic jab about his "real food," and somehow connecting that with me being an American.

    I think he was trying to insult me, but what came through instead was his humanity. He seemed to take pride in his food, which turned out to be quite good. He seemed like a character in a play. Had it been a musical instead of life, he would have sang a song with choreography.

    I went back to that little sandwich place several times after that, cheerfully bearing his insults each time. I called him the Cantankerous Man. For those of you intending to visit Amsterdam, Cantankerous Man's cafe may be found right next door to the Anne Frank museum, along the canal. Go ask him if he serves real food.]

  11. jellybean Says:

    Could it be that the “sales guy” was using a series of buttons with phrases, like with a Soundboard? That might explain why he would say the exact same sentence, but used a human’s judgment to sort out when to use which phrase?

    If so, I think they need a “Yes, I’m a human, but I only have a series of phrase buttons” button. :P

    [James' Reply: I have since had several interactions with WebGreeter technology. Each time except the first time I was able to get the "greeter" to type something that appeared fresh (due to grammatical errors, brevity, and generally appearing responsive to some specific element of what I had just written). Just as in "real life" a human always has the option to "phone it in" instead of making an effort. Problem-solving behavior, as I wrote long ago in my article "Enough About Process, What We Need are Heroes" is a heroic act.]

  12. Bubba Says:

    I’m reasonably sure that Apple’s sales support has the same sort of tech behind it. I had the same difficulty believing that it was a person, to the point where I was disconnected from chat several times because of my attempts to make them go off script

  13. ryan Says:

    i’m going to go with what a previous commenter has said, that typically service posts in businesses are manned by low end low paying low expectation staff or even outsourced to foreign nations on the cheap. in such a situation, the canned responses are exactly what must be used. deviating from the script will lead you into unexplored territory where you might actually have to make decisions that matter, or worse be forced to pass the call up to someone paid more than you which would be the equivalent of not doing your job properly

    [James' Reply: There is already a technology for that. It's called a text file. You put the letters "FAQ" at the top of it. What I'm complaining about is false advertising created when some script-follower POSES as a problem-solver. That's fundamentally the issue, whether in testing, or in customer service. It's misrepresentation of service.]

  14. Ben Simo Says:

    The response from Zach has me laughing. Is it really from webgreeter.com?

    The response is exactly what you’re talking about. It is a canned response that has little to no connection to what you stated. I tried chatting with them this morning. They didn’t question my telling them my name was Rumpelstiltskin Grimm.

    The closest I got to a human response was in response to my response to being asked if they could ask a question they were asking by asking if they could ask. :) I told them they asked the question and gave them an answer. The response was that they hadn’t asked me the question before. I detected some attitude in the response but at the same time the response led me to believe that what I said wasn’t intelligently processed.

    Many companies seem to view this kind of robotic interaction with customers, and potential customers, to be the professional way to communicate. I disagree. Dialog is not dialog when those involved are following scripts like computers.

    How dare we think? How dare we emote? How dare we interact like intelligent people when we could behave like machines? Ugh!

  15. Umar Says:

    totally bullshit..it is an excellent service

    [James' Reply: Um. Would you like to expand on that? Give some reasons? Come on, I'll let you make your case on my website, but you need to use your words.]

  16. joe Says:

    I talked to Jenny at WebGreeter. Personally I can’t see any benefit of using this company. You never get to meet them and you have to train them too. It’s far more cost effective to hire your own chatters locally.

    [James' Reply: That's what I was thinking.]

  17. Rider Of Giraffes Says:

    It’s not necessarily the people or the company at fault.
    Many companies force employees to stay on script in case
    they offer or promise something the company is unwilling
    or unable to honor. That way lies a lawsuit.

    [James' Reply: Don't put a "live operator" icon with a picture of a pretty girl on your site if you can't come through with "nurturing non-threatening helpful fellow human" level of service. That's a misrepresentation deserving of a lawsuit, right there.]

    IANAL, but I do run two companies, and legal advice we
    have received says exactly that. Let your employees go
    “off-script” and run the risk of multi-million dollar
    exposure.

    [James' Reply: I also run two companies. I haven't yet been sued for speaking human-to-human, or allowing my employees to do. For that matter, American Express and CenturyTel have humans who have talked to me recently in a friendly and obviously responsive way and helped me.]

    Don’t blame the employees, don’t blame the companies.
    Blame the litigation culture.

    [James' Reply: It's not a litigation culture, it's FEAR of litigation. That fear is far more of a problem than actual litigation. Or are you saying you actually have been sued for speaking like a person?]

    PS: This is my third attempt at posting a comment.
    First your page crashed my browser, and in neither
    of the first two did the CAPTCHA appear. Your site
    has failed the accessability test for the disabled
    using old hardware.

    [James' Reply: Nuts. I'm sorry. This is WordPress. I wish I didn't have to use this captcha thing, but the spam is overwhelming. You can always email me personally with your comment and I will post it.]

  18. Dave Smith Says:

    Many years back I was involved with a system that was on its way to being something like WebGreeter. The trend at the time was towards “live” systems that started with a bot (backed by some simple, frame-based AI), escalating to a human operator if/when the bot ran into problems. Even then, the operator would access to a bunch of canned responses (in part to guarantee correct spelling!). It was seen as the next step in call center evolution, with humans monitoring dozens of bot-driven conversations. The result can be rather disconcerting.

  19. Rafe Furst Says:

    A friend of mine got me a T-shirt a few weeks ago that says \Why do I get the feeling you are about to fail a Turing Test?\ Maybe he was inspired by WebGreeter…

  20. John Smith Says:

    Why do you make it sound like it’s the end intelligent thought as we know it with your witty remarks? Why are you lashing at everyone for with opinions and rejecting all reality except your own? Are you, inevitably, -that- lonely?

    Congratulations, you’re so human you deserve a gold medal for being the #1 true human on this planet.

    [James' Reply: Usually I don't allow anonymous remarks in comments. After all, I'm not anonymous. I have to pay, in the coin of reputation, for what I choose to say. Some of what I say I later come to regret.

    I'm allowing this comment because it sounds to me like you're hurting, and I'd like to respond to you. But I don't understand what your point is.

    WITTY: I mean, sure, I try to be witty. That's allowed, isn't it? Is there a blog FAQ I don't know about that bans clever turns of phrases?

    REALITY: I don't think I'm rejecting "reality." No one has direct knowledge of reality. Rather I'm arguing in favor of a particular interpretation of reality. That's allowed, too, isn't it? In this Internet culture you can offer opinions on your blog and I can offer them on mine. If I comment on your blog, I expect you to counter-comment or moderate as you see fit, just as I do. It is a nice, free, and symmetrical arrangement.

    I think I "lash" at people who've earned it. Some of my colleagues step in, once in a while, and tell me I've gone too far. I respect colleagues who are willing to perform that service for me. Perhaps you are willing to perform that service. If so, a way to do it so that I'll listen is to provide specific examples and suggest alternative ways I could have gotten my point across.

    In general, I run a rough blog for people who appreciate the world view I live by. There's a place for that in this industry.

    LONELY: I don't understand what loneliness has to do with this. I don't feel lonely. I feel exuberant and gregarious and occasionally irritated. Are any of those words synonyms for lonely?

    HUMAN: Thank you for acknowledging that I'm human. Does that mean it's okay for me to exhibit qualities commonly associated with humans? For instance, when I write, I am expressing something that moves me and hoping to move others. That's a human thing. I'm not in any way setting myself apart from the rest of my species by doing this, except inasmuch as this: I FIGHT BULLIES. So, I set myself apart from them. I push back against the people and institutions that wish for you and I BOTH to shut up and go along with their plans, regardless of what's right or good or useful.

    Anyway, if you are commenting anonymously because you perceive yourself as weak and vulnerable to public humiliation at my hands, then you needn't worry. I withdraw anyone's comment at their request, and I can and have withdrawn whole blog posts against people who felt hurt or wronged by what I've written, and wrote me a respectful letter asking for a retraction.

    I've disclosed my agenda: I fight the bullies who make the world unsafe for good testing and good software. I make my living through such fighting. I feel that in doing so I'm making a better world.

    Now, "John Smith", what is your agenda? Why not disclose that to us in a non-anonymous post?]

  21. Vietor Davis Says:

    As one of the co-founders of a small business, this sort of thing is very frustrating. We take customer service pretty seriously, and run a web chat widget on our website. Use the chat and it’s basically guaranteed that myself or one of the other founders will answer it. It’s chat, it’s not as if it is a huge time commitment or distraction, and it’s been a very valuable sales tool for us.

    I cannot imagine it would be more cost effective in the long term to rent it out to glorified FAQ readers rather than hiring and training some dedicated internal support staff if the volume becomes unmanageable. Sure the per-person expenses will be higher, but does a script reading chat-drone really provide any positive return, both in terms of new customers and in existing ones retained?

    Maybe I’m wrong. We’ve only been around a few years, certainly have not yet grown to mega-corp size, and have a business model that requires long term customer satisfaction, not just a quick buy.

    [James' Reply: Folks, at 10:39pm I decided to see if Vietor at http://www.zettabytestorage.com was telling the truth. I went to his site and activated his live chat. A moment later he replied. This is a slightly abridged version of our chat:

    James Bach: Is this live support?
    (Call accepted by Vietor. Currently in room: Vietor, James Bach.)
    Vietor: This is
    James Bach: It's pretty late at night!
    James Bach: you are up this late?
    James Bach: yes, I was just checking it out
    Vietor: I am, we're in California
    James Bach: I wanted to be able to say something nice for a change!
    Vietor: And tend to work late
    James Bach: ah
    James Bach: well, geez, it seems like it doesn't take you long to type a response
    Vietor: I think it's great that you've verified the claim
    James Bach: and you seem unafraid to be "sued"
    James Bach: excellent
    James Bach: I will go finish moderating your comment.
    James Bach: thanks man!
    Vietor: Any time.
    Vietor: Thank you for checking in

    That's what I expect from live chat support. What is so hard about that?]

  22. ryan Says:

    yeah, i ran into this phenomenon on hp’s website. i was doing a live chat asking questions about a few of their laptops, and though i could tell it was a human being on the other end, my questions were never answered specfically, and i had to repeat myself over and over to get the info i wanted, rather than the sales pitch they were trying to give me. Almost every question i asked, instead of giving me an answer, they would give me a link to a page on their website tht didn’t answer my question. I don’t think i’ll ever use live chat again if i have a choice, and if i ran a website, i’d just use real people on the phone.

  23. The WebGreeters Says:

    [James' Reply: I do not normally accept advertisements in comments, but in this case, it contributes to the discussion.]

    The webgreeter service, as the name implies, greets visitors on a website. The requirement for such a service stems from the need to provide a personal customer care touch over the heretofore impersonal world of the WWW. WebGreeters provides live support services which includes chat software and the trained personnel (also trained for the specific website and its business) to websites, and manages and coordinates a team of greeters (much like a call center) for these websites.

    Make no mistake about it, when chatting with a BOT you would easily know.

    [James' Reply: As you read from my initial post, I did NOT know. It has generally been my experience that the operators communicate mostly in terms of cut and paste, and in the first interaction I had with the service, the webgreeter refused to type natural text to me, while providing "answers" that seemed not to address my questions. You probably have the exact transcripts of what I wrote and what your greeter wrote. I will post those transcripts here for my readers to judge if you email them to me.]

    Ask questions about world news, state that you want proof of the human behind the chat screen or something similar. The following is a quick overview of how the service works and the tools including ‘canned messages’ that are normally used. Remember, GOOGLE recognizes webgreeter services as being customer support rather than ad-BotWare, with the condition that there is a delay before a visitor coming onto a website is actually approached by a greeter; that is why we greet after a small delay.

    The service:
    • Greeters approach visitors with a standard greeting, asking if they require assistance or have any problems or questions regarding the website; normally they also ask visitors exactly what they are looking for.
    • Greeters then push relevant pages to the visitor, directly providing the information required.
    • Greeters will also present marketing content and deals on products or services that the visitors are interested in, thereby filling a sales roll as well as the greeting role. This role is assigned to them by the website they work for.
    In addition to the fact that the greeters use applet based chat windows to converse with clients, they have the following tools at their disposal:
    • Canned messages: These are the automated sounding pre stored messages and are basically meant to;

    o Control spelling mistakes or typos that could otherwise stir up confusion.
    o They are designed so as to free all parties of legal liabilities or misinterpretations.
    o They are designed to prevent discrepancies and to provide standardized information.
    o Save typing time as our greeters are meant to multitask several chats (making them cost effective), and this makes their work easier.

    [James' Reply: The problem with them is that they are the antithesis of human contact. Pre-written messages simply do not convey the message that a human being is with me. Now, if you really want to push pre-written text, there maybe ways to do that without presenting it AS IF the live human typed it in that very second.]

    • ‘Push page’ feature: This feature allows our Greeters to directly push a page of the website directly to open onto a visitors browser and is meant to present specific material of the visitor’s interest.

    Regards,

    Howard White
    Marketing Manager
    WebGreeter.com

    [James' Reply: Thank you for being willing to comment on this thread!]

  24. robert papesch Says:

    Hi James, I met you in Christchurch, New Zealand a couple of years ago. Fascinating talk… anyway this blog on cognition, culture, and computer programming identifies two modes of thinking; “Mapping” is akin to the creative and human-relational; “Packing” is a lower-order level of simple procedure-based thinking. Quote:

    Packers have a whole proceduralised culture that provides behavioural tramlines for just about everything. It’s so complete you don’t even notice it until you solve a problem perfectly effectively one day, by a method that’s not on the list. It might be trivial.. Mappers hardly ever get the upper hand on these cultural issues, but when it does happen it can be hilarious. A packer gave a dinner party and it so happened that over half of the guests were mapper types, IT workers and others. The host pulled a pile of warm plates from the oven, and started handing them to the guy on his left.”Just pass them around!”, he cried cheerfully. Everything went well until he realised he needed to shout “Stop!” [everyone was just passing plates around randomly in an endless loop!] Mappers don’t have a general cultural context to learn from, so we are almost entirely self taught.


    [James' Reply: I haven't heard this distinction between mappers and packers. I'll look into it. Thanks.]

  25. Danny Faught Says:

    This is the second reference to webgreeter.com I’ve seen lately. So I took a look on my Mac, with Firefox. Waited half a minute, no greeting, just a standard web page with a “Live Operator Available” link at the bottom that doesn’t appear to be clickable until I try to click it. Surely it’s not Windows-specific? I tried Safari, which is sometimes compatible with a web site when Firefox is not. Then I got distracted.

    Several minutes later, a Safari window popped up with the requisite greeting. I told her that I was surprised that it took a while for them to contact me, and the answer was a human-sounding “Sorry about that”. I sent a question, and the window then immediately disappeared without a trace. Perhaps I was the victim of a software bug in addition to possible flaws in business process.

    [James' Reply: Another one of the creepy things about their service is that they control the window, not you. It disappears whenever they decide that the conversation is over, and you can't copy the contents of the window.]

  26. Michael M. Butler Says:

    “…is meant to present specific material of the visitor’s interest.”

    Hmm. I wonder if Howard White actually wrote this, or if he had it written for him? Because “material of the visitor’s interest” doesn’t sound like a phrase I’d hear a native speaker of English utter. Not bashing, just remarking.

    Pushing an entire web page at someone is a pretty lazy way to communicate, which I think underlies your point, James. The attitude seems to be:

    If it’s in there somewhere, great. If not, I did what my boss told me to do.

    Especially infuriating if I, the user, have already read the page in question, and told the alleged live helper that.

  27. Michael Bolton Says:

    Here’s a test: When someone asks “And how can I help you today?”, try replying, “Let’s start by promising that neither one of us is going to say anything from a scripted process.” (Uh, naturally you’ll have to modify this for your own purposes each time.) The answer that you get to that request will reveal oodles of information about the company and its approach to managing people.

    —Michael B.

  28. Michael M. Butler Says:

    There’s a company I won’t name who basically invented the “computer-on-a-stamp-sized-board” as sold to hobbyists and engineers for tinkering and prototyping. They have open telephone support starting at something like 7:30 AM Pacific time. I called them a few months back and the person I talked to gave his name as {nickname}. He started to handle things in a formulaic “first let’s check…” way, but when he realized I knew I had an “outlier” problem, he stepped up immediately.

    It was three minutes after that call was completed (with 100% customer satisfaction) that I realized {nickname} was the CEO, up early and taking tech support calls just to stay connected to the marketplace.

    How impractical and inefficient of him. /sarcasm

    [James' Reply: You can name the company if you want to. I happen to know you are NOT merely shilling for them.]

  29. Kirsten Says:

    Sorry, I haven’t been designed to discuss human. How can I help you?

    I share my experience with PayPals virtual Agent, Sarah:

    Paypal sends in agent Sarah to help me sort out my ebay problem. Here is the dialogue. I must admit that I did not initially read the first sentence explaining that she was not human. So, my first entry overwhelmed her. Although, simplifying did not make it any better. To top it off and add insult to my frustration, when I did speak to a real person, she was laughing about the this diaglogue with a coworker and said that Sarah should be a blonde and not a brunette.

    Here it is, both sad and hilarious. Thanks PayPal:

    Sarah – PayPal :
    Hello, I’m Sarah and I’m here to answer your questions about PayPal. Even though I’d like to be a real person, I’m not. I’m programmed to answer your questions. How can I help you?
    You :
    I am an active ebay buyer. I usually choose to pay for items over $100 with AMEX; however, after reading PayPal’s security guarantee with each transaction, and having to press Yes to Change, when presented with the ubiquitous “Are you sure you want to use a different payment method?”, I’ve relented and relied on direct withdrawal from my bank account. – the obviously preferred method by PayPal. Now I am experiencing a potential problem with a seller (they sent the wrong item). Paypal’s resolution center says that refunds are handled outside of Paypal. If I had charged the item to Amex, I could ask for a hold or refund. Please explain to me how Paypal protects me when I pay with direct withdrawal?
    Sarah – PayPal :
    Sorry, but your phrase is too long and I will not be able to perform a full analysis. I would be very grateful if you could try to make it a bit shorter and more concise.
    You :
    So Sarah, you are computerized and not a real person?
    Sarah – PayPal :
    I’m listening. How can I help you?
    You :
    Please explain to me how Paypal protects me when I pay with direct withdrawal and do not receive the right item from the seller?
    Sarah – PayPal :
    Sorry, but your phrase is too long and I will not be able to perform a full analysis. I would be very grateful if you could try to make it a bit shorter and more concise.
    You :
    Please give me a phone number so I can talk to a live person.
    Sarah – PayPal :
    If I can’t answer your question, you can visit the Help Center, or contact us by email or telephone. The customer service number will vary, depending on which account type you have.

    To find the phone number:
    1. Click here for the Contact Us page.
    2. If you haven’t already done so, log in to your PayPal account.
    3. The PayPal Customer Service phone number for your account is listed.

    To contact us by email:
    1. Click here for the Email us page.
    2. Log in to your PayPal account.
    3. Provide the requested information, and then click Continue.

    Before we continue, was this helpful?
    You :
    No
    Sarah – PayPal :
    I’ve just opened the right page for you.
    Sarah – PayPal :
    Anything else I can do to help you
    You :
    If you were a live person, perhaps. But your lack of analytical abilities is not helpful.
    Sarah – PayPal :
    I’m sorry but I’m unavailable right now. Please try again in a few minutes.
    You :
    Circuit overload?
    Sarah – PayPal :
    I’m not sure I’ve understood. Could you please rephrase your question?
    You :
    Thanks for great blog fodder
    Sarah – PayPal :
    You’re welcome. If you think of anything else I can help you with, I’d be happy to assist you.
    You :
    Tip, be sure to turn yourself off before you take a bath
    Sarah – PayPal :
    I apologize if you can’t find the information you seek by speaking with me. Please enter a new question, visit the Help Center, or Contact Us.
    You :
    Hopeless
    Sarah – PayPal :
    Really, that frustrates me… Because I do my best to improve every day. If you think of anything else I can help you with, I’d be happy to assist you.
    You :
    Bring a human in to assist you.
    Sarah – PayPal :
    Sorry, I haven’t been designed to discuss human. How can I help you?

    [James' Reply: I think this last question is actually intended as rhetorical... Indeed, not being designed to "discuss human", how CAN it help you?

    It's creepy that it's programmers give it the name "Sarah." Are we supposed to make allowances for it because they plaster a sign of feminine character on it? I think that's what they must think.

    Microsoft tried that with "Microsoft Bob" some years ago. That, too, was a dismal failure.]

  30. Rusa X. Says:

    “Kevin- The fact that they are easy and repeatable and effective make them usable by any idiot. Idiots are cheaper than smart people, hence the reason that you build business processes.

    [James' Reply: Any idiot? Any? How could you make such a claim? What evidence can you offer?]

    Kevin – The fact that McDonalds can make the same hamburger in a predictable and efficient way comes from a business process.”

    Although I am not Kevin, I believe this was reference to Joel Spolsky’s excellent article – Big Macs vs. The Naked Chef. I think you should read it because it concerns the ~decreased~ quality of anything when talent/ humanity is ‘canned’ and the article is entertaining. Laughter is good for people. (No, I have no suppporting evidence for that with me right now.)

    It is located at http://www.joelonsoftware.com/articles/fog0000000024.html

    [James' Reply: Thanks for the reference.]

    Cheers,
    Rusa

  31. Ignatios Souvatzis Says:

    I wanted to share my experience of failing the Turing test – I was subjected to ReCaptcha by some website that requires it for feedback – but I see that you require ReCaptcha solving also, so I’ll most probably fail here, too.

    In case not: look at my \website\.

  32. Oliver Smith Says:

    ” It was, for him, a test of whether a machine can think”

    Actually Turing in his original paper from Mind, didn’t call it a test, but an imitation game. This is an interesting distinction from the man himself. Which is one reason why I think that people have taken his thinking outside of the true context.

    Blay Whitby wrote a really good paper on this:

    http://www.cogs.susx.ac.uk/users/blayw/tt.html

    [James' Reply: It was a test, whether he called it one or not.]

  33. Michael M. Butler Says:

    OK, James–at your request, I’ll name names. The company is-was Parallax, Inc., and the guy is Chip Gracey. Not sure what his exact position is, it might not be CEO, but he’s one of the founders; so in a sense it’s like having Woz answer the phones at Apple.

    Cool, in my book. And the antithesis of botpeople.

  34. Dave Says:

    This service managed to piss me off, something that no chat-bot has ever managed to do.

    I posed as a genuinely interested customer, throwing a few curve-balls every now and then
    to test the water. I was able to tease out many custom responses (rather than scripted ones),
    eventually getting the operator to say:
    “with that attitude i doubt that you even have a job” after asking it on a date.

    I feigned offense, explaining that with such a new and foreign technology, users could be
    expected to be suspicious and ask odd questions, and under no circumstances would I be using
    a service that would ever insult one of my potential clients… especially with a response that could
    be so demoralising in the current economic climate… and was disappointed that the operator
    closed the window before i could save the conversation.

    I had a second attempt at talking to the operator, some moments later, found here:
    http://dave.icy.com.au/webgreeter.gif

    I thought I’d share my experiences :)

  35. lvs Says:

    ha ha ha. this is hilarious. Anyway after Watson beats humans at Jeopardy I guess this wont matter?

  36. Max Says:

    I StumbledUpon this website. And I read a little so I went to the link at the top. Thought I’d play around with it. You know, do a little claims verification. What I got back surprised me.

    Daniel: Hello, I am the website greeter. Welcome to WebGreeter.com. How may I help you?
    Visitor2863: not a thing. just curious
    Daniel: Let me know if I can be of any assistance to you.
    Visitor2863: are you a human being, or a chatbot
    Daniel: a human, but hey, wouldn’t a bot be saying the same thing?
    Daniel: ah well, anyways.. nothing i would say would change a rugged mind.. are you one? :P
    Visitor2863: no, no. i believe you
    Visitor2863: you are human. and i hope that at the end of your work, you go home, relax, take it easy and have a good time
    Daniel: good that you do sir, so i pass the turing test? :)
    Visitor2863: you did, you sly dog, have a good one.
    Daniel: Thanks bunch! I hope you have a good and safe weekend as well!
    Daniel: rotfl!

    [James' Reply: Sounds like a jerk, to me. What the heck is a rugged mind? Does he mean an appropriately skeptical adult mind? Is he criticizing you for wondering if you are being manipulated by a script? I would think, by now, they would have a more polite and graceful way of answering that question.

    The main thing that made me think I was talking to a chatbot was the unresponsiveness of the guy on the other side. I asked questions that were ignored.]

  37. Henry Says:

    Maybe he did not know the answers to some specific questions you asked and would have wanted to refer you to someone, i know this as I stumbled upon this website sometime ago too and found it to be immensely helpful, not robotic at all… no one makes a chat bot that smart, not even microsoft :P

    [James' Reply: This is a hazard of their technology that they must manage better. His answers to my questions were simply not responsive to my questions, and gave no hint of human understanding.]

  38. Luthien Says:

    Hi,

    I’ve been wondering about the same thing after using the support “live chat” of Fatcow.com (an internet provider). It leaves me with a very unsatisfied feeling of not being able to determine whether I was talking to a real person (because of their saying-so) or to a chat bot (because they “felt” like a bot). And whether it’s a human chatting to me or not, I find it unsettling either way. In the one case, it’s scary how incredibly “dehumanized” they can make people act: I find it outright spooky. In the other case it’s scary because they are lying.

    I’ve spent some time trying to figure out what the case is, but somehow no-one answers my questions. I’m definitely not into conspiracy thinking, but I cannot wholly dismiss the feeling that if you would want to suggest something like that, you cannot do better than these “liveperson chats” actually do.

  39. Live chat agent Says:

    Don’t you think that sometimes live agents may just get frustrated by time wasters calling to ask the agent on a date? Is it really funny for someone older than 10? Is it human behavior? If you are a human, you treat others the same way you want to be treated yourself. A human being should be able to understand that the agent may be helping several other customers while you are here just for fun.

    [James' Reply: It sounds annoying, but yes, actually, that IS human behavior. This is what you sign up to handle when you take that job. But that has nothing to do with my post, does it? I wasn't trying to harass anyone. I was trying to get my questions answered. In the process, the technology being used, combined with the decisions made by the agent, gave me the strong impression that the live agent was not live at all.]

    James, I understand that it wasn’t the case with your conversation with WebGreeter, but I’m so disappointed by trolls in this thread who started calling to different agents in different companies just to provoke and offend those people, ask them on a date, etc. You will never understand how upsetting those pranks may be be unless you have to deal with them every day.

    [James' Reply: I suppose in some ways you get used to it and in some ways you get sick of it. But in the case of the people on this thread, we are testers. We are not doing this to annoy anyone, but rather to understand the technology and service better. It's important for our education and the good of the industry.

    There is another element involved, too. It's not your fault, there are certain ethical problems that come along with combining chatbot technology with human communicators. Blurring the line between machine and human is dangerous for all involved. It creates the conditions where customers can easily feel abused and, as you point out, agents feel abused, too. Just the term "live agent" makes me a little sick. You are not a "live agent", god dammit, you are an agent. There is no such thing as a non-live customer service agent.

    I'm amused to hear that lots of people who read my post went and tested the chat system. It's a strike back at the designers and managers who promoted the dumb idea of cut and paste tech support. Having said that, I feel bad for the poor people (I won't call them l*** a*****) who are in the middle between their foolish callous management and their clever callous tormentors.]

    I’m convinced that anyone may say something stupid or not understand something correctly under the emotional pressure, if the person who asks questions has a purpose to piss you off. So, just for the purpose of staying professional, protecting own nerves and saving the time of other customers the agent may be chatting with, he/she has to become “bottish” with trolls. Otherwise the agent just won’t have time for solving real customer’s problems.

    [James' Reply: That's an interesting test result.]

    At the same time I must admit that a live chat agent must be professional no matter what. And even if you think “boy, does this person have a job if he/she has so much time for this bullshit?”, you should never type that :)

    [James' Reply: Agreed.]

  40. Jay Says:

    I didn’t actually read every response to your article so I’m probably repeating but this type of sales is exactly why people have zero patience with sales teams of any type. The reason we hang up as soon as its clear that a phone call is sales based, or how we are always “just looking” at any retail store, all stems from aggressive sales people that don’t care what we buy, just that we buy from them.

    On a side note, you should check out Searle’s “Chinese Room”, it basically denies the possibility that the Turing test could ever prove that AI is actually “aware”

    [James' Reply: Searle's thought experiment is one of the great challenges in modern philosophy. However, there have been interesting replies to it and the matter is not clear cut.

    One of the more powerful critiques of it claims that Searle's example is incoherent on its face, because it is logically impossible to construct a set of rules for speaking Chinese that would satisfy the requirements of his Room. Speaking any human language requires non-algorithmic behavior-- behavior that we cannot reduce to explicit rules of any kind, even in principle. See Collins book Tacit and Explicit Knowledge for a clear exposition of that view. He refers to Searle.

    I certainly agree that following rules to do ANYTHING is not that same as problem-solving, and therefore involves a lower degree of awareness and much less resilience and reliability in the face of trouble. Context-Driven test methodology is all cultivating that higher awareness, which we simply call "skill."]

Leave a Reply

Comments are moderated. You might want to read my commenting policy before you make a comment...