Interview with Lina Wolff
If artificial agents – robots, chatbots, and various AI systems – appear as humanlike, it may be easier for us humans to interact with them. And those who design artificial agents have many opportunities to try to achieve such likeness. One possibility is to give them a natural language – a language that humans use in speech or writing (as opposed to, for example, a programming language). This possibility is of course already used: Siri in the Apple phone has a voice, and ChatGPT can respond with text that often appears to be written by a human. But is it possible to make an artificial agent even more humanlike, and more like certain types of humans, by trying to influence how it uses its language? If this is possible, what should those who want to explore this actually do? A professional category facing such questions is fiction writers who need to create believable fictional characters and to provide them with specific traits. Perhaps it is precisely writers that designers of artificial agents should try to hire to refine the agents’ language. I had the opportunity to discuss this with Lina Wolff, who has written the novels Bret Easton Ellis and the Other Dogs, The Polyglot Lovers, and
Carnality. Her most recent novel is The Devil’s Grip.
Magnus: Many believe that we will soon communicate more and more with various machines. And many of these machines will initially communicate primarily with their language, either only with text, like ChatGPT today, or with a synthetic voice. Assuming thata person’s language says a lot about who the person is – if the person is a real human – one would also assume that there is an interest among those developing a communicating machine to somehow influence its language. And perhaps they will seek help from fiction writers. If
someone comes to you, and says “You, a fiction writer, who can immerse people in your stories through language, can you help us with how our robot should speak?”, would you find it strange? Or would you see it as a challenge?
Lina: From my perspective, I understand that we are facing such a development; this will become more and more commonplace, and we cannot escape it. If it were a commercial product that helps people in everyday life in a way that is beneficial to people or humanity at large, well, then I don’t see why I wouldn’t be part of something like that.
Magnus: Let’s assume that you have said yes to such a task, and it concerns a home robot that is designed to help people. It can listen, and it can speak, but it should be a bit more exciting than that: it should not just repeat what someone has asked it to do, and it should not just confirm that it has done what was desired. It should feel a bit more alive, a bit more like a human. Given this, I would like to ask you a few things about your view on the credibility of characters you create in your stories. Is credibility central? Must a fictional character have
certain traits to be perceived as real? Is this something you think about? Or is it somehow taken for granted that the reader, who knows that fictional characters do not exist for real, would engage in suspension of such beliefs to be able to enjoy a story? If the reader is indeed doing this, maybe the author does not need to put in so much effort into make characters credible?
Lina: Credibility is very central when writing fiction. It is about tuning in to the right register and conveying it in a way that works in the story, but above all, it is about using dialogue to make the reader create a clear picture of who is speaking. When writing fiction, you also have
many other means of characterization than what a character says: you can, for example, describe facial expressions and gestures. But dialogue is essential for credibility. It is a lot about attitude. You have to charge a character with attitude, and you can do that with language. If, for example, you want to create a specific personality for a character, you can make it laconic or let it speak in short sentences or keep it silent. And if I want to create a character that I like, I give it an ability to listen and not talk in overly long harangues.
Magnus: Often, we like people who are trustworthy. Is that something a writer can work with?
Lina: There is a scene in The English Patient that is very telling in this regard. It is when a man and a woman are driving in the desert, the car overturns, and then a sandstorm comes and they get buried. “Do you think we will get out of here?” she asks. “I absolutely think so”, he says. Then she says this: “If you hadn´t said "absolutely" I would have believed you. This, “absolutely”, is exaggerated emphasis in this context. I have used that many times. It gives a feeling that something is wrong when something that does not need to be said is said anyway.
Magnus: Characters in fiction often speak with correct grammar and they use a proper word order. I brought out one of your books, Carnality, and in the beginning there is a character who initially you do not get to know very well. This character, “she”, is quite concise when she speaks, but she still always speaks in complete sentences with the right word order. Is it important for you that it should be like that? Even if characters are a bit strange and imperfect, should their language still be correct? One could imagine that one way to create a credible human character is to let it speak in the same incomplete way that we humans often do in
everyday situations.
Lina: I think that a correct language gives a sense that the thought behind what is said is clear. And this clarity is important in relation to the reader. There is no reason to give characters an incorrect language unless you want to portray some kind of dysfunction. I think that even the anguage of robots must be correct; I would not want to communicate with a robot that does not express itself correctly, because that would give me the feeling that it is misprogrammed. At the same time, much of our communication is non-verbal. We are good at reading both body language and vocal tones. So if you add this, it is not certain that a robot that speaks perfectly is very good if it cannot communicate non-verbally. I think it would be incredibly scary to sit in front of a perfect interlocutor who does not communicate on the non-verbal level. It gives double signals, and this is perhaps the least trustworthy thing in human communication. That is, saying something verbally but signaling something else.
Magnus: Let’s move on to emotions. Suppose you want the reader to understand that a certain character is in a certain emotional state without letting it say this explicitly. If a character is supposed to be happy, you don’t want it to say “I am happy”, and you don’t want a narrator’s voice to say “he is happy”. You want the reader to understand this with more subtle means. Is there anything a writer can do in this case? I ask about this in terms of a robot that cannot feel any real emotions, at least not yet, but the designer wants it to appear as if it indeed has the capability to experience emotions and wants to convey this by the way the robot speaks.
Lina: Yes, you can show characters’ emotional states without being explicit. Let’s say, for example, that a ceiling is leaking. And you want to convey that someone is very happy that the leak is fixed. Then you can let the person say “Finally, the roof is fixed! We don’t have to
lie awake tonight thinking about this anymore”.
Magnus: Would you feel that it’s a bit of cheating or ugly to explicitly write that a certain character is happy, sad, or angry? Is it bad style to use such adjectives like these to describe someone’s state of mind?
Lina: It becomes a bit rudimentary. It becomes a bit too simple. Of course, you need adjectives, but you have to use them in a way so that the reader understands that it is a character’s view of the matter. They should not come too often from an omniscient narrator, at least not if you want to try to achieve some kind of objectivity in the storytelling that contrasts with the characters’ subjective view of the world. Hemingway’s old advice is highly applicable: Show, don’t tell. Don’t tell that someone is happy. Let the person instead walk with a light step or hum... But I do not know if this can be applied to a robot without being terrifying.
Magnus: Another thing I thought about when reading your book Carnality has to do with the person, “she”, who is introduced at the beginning. It takes some time before you get to know her. It does not happen immediately, and she is also quite brief in her speech. Is it a deliberate choice by you that a reader should not immediately know everything about a character?
Lina: It is impossible to immediately communicate everything about a person if you want them to be complex. I also often want to portray some kind of change or process, and then it is not possible to say everything at the beginning. And at the beginning of a story, it is a lot about getting the reader on board; you risk exhausting the reader if you always keep the text at a very intense stage. In the case of “she” in Carnality, I also wanted her to be a bit inconspicuous to highlight other characters who come in later in the story.
Magnus: If you want to simulate the presence of a real person, who you can never get to know completely immediately, is there a point in not revealing too much about a person at the beginning of a story?
Lina: For this character, “she,” there is a progression that must take place; it ends with her becoming, in a way, another person. I would almost say that the essence and heart of Carnality is her progression. And then it becomes important how you present the person over
time: you have to be very careful about how you release information and when the reader should learn different things. The reader never knows more about the story than the author knows, and a good novel must work in that way. It may happen that things about a character
somehow leak, in an unintended way from the author’s point of view, but that is beyond the plot and how one organizes voices in a story.
Magnus: What about sympathy for a character in a story? I understand that you can make readers sympathetic to characters by letting characters do and think various things. But if you were to try to build sympathy for a character solely through what the character says, what can a writer do?
Lina: I do not have a specific formula for that, but spontaneously I think that the character’s language should signal security and clarity. It should not be self-glorifying, no boasting. The character should listen more than it speaks. And other basic interpersonal rules should be followed. The character should not laugh at the wrong time; if someone tells something painful, the character cannot start laughing in the middle of it. It is also difficult to like someone who is unclear about his or her own boundaries, and it is difficult to like those who makes accusations or provide explanations about why another person is doing whatever he or she is doing. Nowadays, this may be a bit complicated because we are, both as writers and readers, too seasoned to believe that “pure heroes” exist. Therefore, as a reader, you can
develop sympathy for creaking personalities, and as a writer, you are, of course, aware of this when creating a character. But to develop sympathy in such cases, you must sense some kind of vulnerability in the character.
Magnus: What you say about self-glorifying is interesting. Have you ever been on LinkedIn?
Lina: No, I have not.
Magnus: I am a beginner there, and sometimes you get amazed. I did not think it was possible for adults to say such distinctly positive things about themselves.
Lina: A good rule of thumb is probably to avoid talking too much about oneself in overly positive terms. The listener is – even on an unconscious level – fully aware that he or she is
the one who determines the impression that others make. If someone says that he or she is likeable, I probably would not blindly believe it. I will weigh in all other non-verbal information available to make that decision. If someone tells me that he is a likeable person, I would probably think that he is kidding himself. Or me. When someone instead talks about how excellent they are, it often indicates a need for confirmation, which in turn is often a sign of not being very strong. Strong people, who do not need to say that they are strong, are often
the ones we may perceive as most likeable, precisely because we do not feel threatened, and also because they induce confidence. And vice versa, if you want to create a character that should not induce confidence, you can let it speak ill of others or use many adjectives loaded with values when it is describing other people. These aspects can evoke a feeling in the reader, just as in people in ordinary life, that it is a character one should be cautious with.
Magnus: Let me also ask about a more specific thing regarding a character’s language, namely the length of sentences when the character speaks. Is this something you think is a tool to signal a character’s traits?
Lina: If I have a character that I want to convey credibility, I would probably make its language concise. No long harangues, especially if the character is not addressed. Then there are stylistically driven monologues where characters talk about themselves without even a period occurring, but that’s another thing, I believe.
Magnus: Many listen to audiobooks nowadays. And when a character in such a book speaks I believe that there are good opportunities to make it sound like real everyday speech – with incomplete sentences, humming, and filler words such as “like”, “sort of” and “kind of”. Can this make a character appear more real? And would it make sense to try to simulate this in a
written text by letting characters use colloquial language?
Lina: I would absolutely not want to remove all my own “like” in my interpersonal communication; it’s a kind of shock absorber in a conversation. But I would not include “like” in a text. If I were to do that, it would be to signal that a person is not trained in his or her speech. And if it is a first-person perspective that should be very close to a certain character with a certain attitude, I might also use “like”.
Magnus: Have you ever included a non-human character that speaks or thinks in a story? It does not have to be a robot, it can be something else – an animal or whatever. A piano that suddenly starts thinking.
Lina: No, I have not. But right now, I´m sketching a story that has a fairly significant amount of artificial intelligence in it. I do not want to say much more about it now.
Magnus: If one were to include a non-human in a story – a non-human that is supposed to speak or think in front of a reader – and if one wants to emphasize that it is indeed a non-
human, can it create extra credibility to incorporate some strange quirks in such a character’s way of expressing itself?
Lina: Ha-ha, I would think so. I would perceive it as more trust-building. If I communicate with a machine, I would find it a bit awkward if I had to sit and be on my guard, thinking “there it didn’t manage to trick me, and there it did”. I would probably feel more confident if I understood that the machine understands that a machine is just a machine, and that we differ as beings.
Magnus: Have you read Klara and the Sun?
Magnus: Have you read Klara and the Sun?
Lina: No, I have not.
Magnus: It is a first-person narrative in which the main character is a service robot, an artificial friend. It talks extensively in various dialogues, and you get to know it through what it thinks. I think that Kazuo Ishiguro manages quite well to describe the inside of a machine that wants to be with people. It must be a certain challenge for a writer to try to portray the inner life of a machine?
Lina: Absolutely. I have thought a lot about that myself. How do you communicate that it is, in fact, artificial intelligence? I'll read it. Thanks for that suggestion.
Lina: Absolutely. I have thought a lot about that myself. How do you communicate that it is, in fact, artificial intelligence? I'll read it. Thanks for that suggestion.
Magnus: I believe that Ishiguro must have thought that he cannot make the robot too human-like because then it can be unclear that the robot is, in fact, not a human. So there must be something that grates and becomes a bit strange here and there. I guess he must have thought so.
Lina: Yes, of course.
Magnus: Lina, thank you for having this conversation with me.