The Future is Conversation

By – Dr. Vinnakota Saran Chaitanya, Assistant Professor, Department of Computer Science and Engineering,

School of Engineering and Sciences,SRM University AP,(Amaravati)

Imagine you walk into a shop because you want to buy something, but you are not really sure what to buy. You have a vague idea in your mind, but you are confused about what would actually suit you. You look around for a while, and the shopkeeper notices that you are unsure. He comes over and asks what you are looking for. You explain what you need. He asks a few more questions about your purpose, your preferences, and your budget. He shows you a few options. You like one, reject another, and then ask if there is something similar at a lower price. The conversation continues, and slowly you become clearer about what you want. You may even walk out of the shop with something that you had not thought of buying when you first entered.

Now imagine if the recommendation system in an application could do something similar. Imagine opening a shopping application, and instead of simply seeing hundreds of products, you could tell it, “I am looking for something for everyday use. I don’t want to spend too much, and I prefer something simple. The system could ask you a few questions and show you some options. You could say, “I don’t like this one. Show me something similar.” It could respond again. You could change your mind, add another preference, or ask why a particular product was recommended. What you have then is no longer just a recommendation. It is a conversation.

This is the idea behind conversational recommender systems. In simple terms, these are recommendation systems that allow us to interact with them through conversation. Instead of expecting us to know exactly what we want and type the right words into a search box, they can listen to what we say, ask questions, and gradually understand what might be suitable for us.

For years, recommendation systems have learned about us mainly by watching what we do. But our actions do not always tell the complete story. They look at what we click, what we watch, what we buy, what we skip, and what we search for. From these activities, they try to understand our interests and predict what we might want next. This has worked remarkably well. It is one of the reasons why the recommendations we see on shopping, streaming, and social media platforms often feel surprisingly personal.

But our actions do not always tell the complete story.

Suppose you click on a laptop because you are curious about its specifications. The recommendation system may assume that you are interested in buying it. You watch a movie because your friend selected it, but the recommendation system may think it is your preferred genre. You search for a gift for someone else, but the recommendation system may treat that search as a reflection of your own interests. A click tells the system what we did. It does not always tell the system why we did it.

A conversation can help fill that gap.

Think about the shopkeeper again. If you tell him, “I need something for a meeting, but I don’t want it to look too formal,” he immediately has a better idea of what you need. He can ask another question and change his suggestions based on your answer. We often make decisions in exactly this way. We start with a rough idea, talk about it, see a few choices, and slowly understand what we really want.

This is what makes conversational recommendations interesting. The system does not have to make one guess and stop there. We can tell it when it is wrong. We can explain what we like. We can reject a suggestion. We can ask for something cheaper, simpler, newer, or completely different. The recommendation can change as our conversation changes.

Consider planning a short trip. You may begin by saying that you want to go somewhere nearby. Then you may realise that you would prefer a quiet place. A few moments later, you may add that you want good food and do not want to spend too much. If the recommendation system understands the conversation, each answer can help it narrow down the choices. You did not need to know your exact preference at the beginning. The conversation helped you arrive at it.

This is where recent developments in artificial intelligence are making conversational recommendations particularly interesting. Computers are becoming much better at understanding the way people naturally communicate. We can explain something in our own words, ask follow-up questions, change our minds, and continue the conversation without starting all over again. Large language models have made this kind of interaction much more natural, and researchers are exploring how these capabilities can be brought into recommendation systems.

But there is a big difference between talking well and recommending well.

A recommendation system may give a very natural-sounding answer and still recommend something that is not useful. It may misunderstand what matters to us or gives importance to the wrong preference. It may also provide incomplete or incorrect information. So, the real challenge is not simply to make a machine talk like a person. The challenge is to make it understand what the person actually needs.

There is also a question of trust.

When we talk to a conversational recommendation system, we may reveal much more than we reveal through a simple click. We may talk about our interests, our plans, our preferences, and the reasons behind our choices. If these systems remember our conversations, they may gradually build a much deeper understanding of us. That can make recommendations better, but it also raises important questions. What should the system remember? How long should it remember it? Who should have access to that information? And should we always know why something has been recommended to us?

At the same time, understanding a user better can open up another interesting possibility. A good recommendation is not always about giving us exactly what we ask for. Sometimes, it is about helping us discover something we did not know we wanted. Think about a shopkeeper again. We may walk into a shop asking for one particular thing, but after talking to us, the shopkeeper may show us something different and say, “I think you might like this.” We may not have considered it ourselves, but after seeing it, we may realise that it is actually a better choice. A conversational recommendation system could do something similar. It can understand what we like and then use that understanding to introduce us to something new that may genuinely make us happy.

This is one of the interesting directions in current research. Researchers are looking at how recommendation systems can understand user preferences, along with changing preferences, remember the context of a conversation, ask useful questions, explain their suggestions, and help people discover options they may not have considered themselves.

For a long time, recommendation systems have mainly tried to answer one question: “What is this person likely to click on next?” Conversational recommendations bring a different way of thinking. Instead of only trying to predict our next action, the system can ask, “What is this person actually looking for?” Just like a shopkeeper who talks to us before showing us something, the system can learn more by listening to us.

This also changes the role we play in the recommendation process. We are no longer just clicking on options while the system quietly learns from our behaviour. We can now take part in the process. We can tell the system what we want, correct it when it gets something wrong, ask why it made a particular suggestion, or simply say, “Show me something different.” The recommendation becomes something we arrive at together, rather than something the system simply places in front of us.

This kind of interaction becomes especially useful when we have too many choices. Today, whether we are choosing a movie, a phone, a restaurant, a holiday, or a course, finding options is rarely the difficult part. We are already surrounded by them. The difficult part is knowing which one is right for us. A good conversation can help us narrow down those choices by focusing on what actually matters to us.

That is where conversational recommendation systems could make a real difference. Instead of simply placing more options in front of us, they can understand what we mean, ask the right questions, listen to our answers, and gradually help us arrive at a choice. The experience could be much closer to talking to a good shopkeeper who takes the time to understand what we need rather than simply showing us everything available.

When we’re drowned in endless choices, the real challenge isn’t finding options anymore. It’s finding clarity. By moving away from endless scrolling and toward actual dialogue, conversational recommenders make searching for something feel less like working a search bar and more like working together. In the end, the future of recommendation isn’t about giving people a longer list. It’s about having a better conversation.