ChatGPT in Tourism
There has been a lot of controversy lately about big language models like ChatGPT, Bard, and especially about their application in the travel industry.
There are many discussions and articles, and in general, these new technologies are evaluated positively.
We, as a company promoting smart tourism around the world, cannot stand aside and want to show our vision of this problem. We tried to figure out whether this technology can bring any benefit to the tourist, as the end user of any travel services.
First you need to understand what ChatGPT (Generative Pre-trained Transformer) is. Such technologies are also called large language models, which better conveys their essence. These chat bots generate the most suitable words based on existing phrases and thus construct sentences. What words and phrases are better to choose they know from the results of preliminary training in real dialogues between people. They simply collect together the most frequently occurring words and expressions, that is, they match the next word to the previous ones with the best probability. Based on this understanding, we express some doubts about the usefulness of this technology in the tourism industry.
Below we will consider several options for using this technology from our point of view.
Building travel itineraries with ChatGPT
The problem is that when creating routes, the chat bot will be guided by the reviews and opinions expressed by the majority of tourists. That is, when choosing a tour, it will suggest the most popular routes, thus contributing to overtourism on the one hand and depriving the route of individuality on the other hand. Most likely the proposed routes will be banal and boring.
Creating tourist route descriptions with ChatGPT
As you know, the tourism industry involves creativity, and even the same route can be described by different tour operators in different ways, taking into account the interests of their target audience. Moreover, good tour operators develop their own unique tours in order to stand out from the competition and provide consumers with a truly unique and inimitable experience. So a good tour operator will carefully and carefully develop not only the route, but also the description for it, because it wants to inspire the tourist to complete this route and will not entrust the creation of the description to the machine.
Perhaps many tour operators will create descriptions using robots, but then these descriptions will be very similar to each other, and those who continue to create manual descriptions will be able to stand out from the competition.
Creating texts for social media posts
Unfortunately, many travel businesses sin by constantly publishing the same type of posts on social networks, believing that the rhythm gives them some advantages over others. But in fact, frequent posting inevitably reduces posts quality. Unfortunately, not everyone understands that the popularity of a page on a social network directly depends on how carefully users read new posts and how many emotional reactions such posts cause in users. The more reactions to the post, the more often it is shown to readers by the social network. The number of subscribers is growing for those pages in social networks that managed to surprise the reader with something. That is why trusting a robot to communicate with potential customers should be done with caution.
Creating valuable content for the target audience
The use of a chatbot for this purpose is also questionable, since it summarizes the content written by people, but there is no guarantee that it will choose the best one. Of course, a robot can write quite a lot of content, but is the number of letters a measure of value? And it should be remembered that any tour organization creates content either for inspiration or to draw attention to themselves or to their product. Therefore, the value of the content is given not by the amount of work on it, but rather by the product tied to it and the brand.
Idea generation for blogs
This, too, can be a problem, as companies run or should run their blogs not just to write something there, but to give their customers additional information. Partly, of course, companies can gain the trust of customers by blogging. But how relevant and useful the content created by the machine for users will be is a big question. And does the content created by the chatbot add the credibility to company ? Rhetorical question…
Answering emails from clients
As a rule, people email the company in extreme cases, for example, when they want to change a booking, or complain about something, or clarify some non-standard services. Moreover, emails are often written by potential customers who have not yet made a choice and have not made a purchase decision. Is it worth trusting a robot for such a responsible matter? If we want to build tourism for people, then we must always understand that a person writes an email because he/she needs support or information and he/she needs attention. Unfortunately, even support services staffed by people often respond according to a template and formally and for a simple question send a bunch of links to the client instead of real words.
Responses to customer reviews
Reviews are a very important factor in choosing a particular travel product and potential customers are guided not only by the reviews themselves but also by how companies respond to them. Therefore, the response to the review should be at least personalized and express gratitude to the person who wrote it. Well, if the review was negative, then the company should sincerely apologize and list what exactly was done to correct the shortcomings and, possibly, offer some kind of compensation for the inconvenience to the client. The robot is unlikely to cope with this, since it does not have specific knowledge and a set of actions for each specific case.
Real-time customer support
A very dubious feature. As we have already figured out, the language model selects the most popular answers and, moreover, is currently trained on data up to 2021. And real-time support is needed precisely in order to give accurate and fast information. The client will call or write to the airline support only to find out how to change a ticket or find out about a flight delay, and not to be asked a lot of questions or beautifully told about what he/she does not need at the moment.
Training new employees
Usually, travel companies teach their people corporate rules, but they tend to convey an individual corporate spirit, especially relations with clients. Staff training supports that elusive thing that makes a brand special in the eyes of the customer. Such learning occurs mainly on an emotional level, which is most likely unattainable for a robot.
Since large language models select the most popular phrases, they are unlikely to be able to adapt to the characteristics of each individual tourist. The recommendations of large language models may most likely produce general recommendations based on the opinion of the majority, which, as we know, is not always right.
This is one of the few things that could be entrusted to the machine, but there is a pitfall – the robot can lose its meaning and idea for the sake of the correctness of speech. This is especially true for something unusual and non-trivial. And this is dangerous because a tourist can receive distorted information and either come to the wrong place, or be disappointed in their expectations.
As we know, language models are not very accurate, they cannot always indicate the source of information, and in research and insiders, arithmetic accuracy is important for decision making. Of course, texts, which are quite numerous in any study, can be written by a machine. But to what extent will these texts and, moreover, conclusions and hypotheses correspond to reality?
Creation of SEO optimized texts
This is a common mistake when texts describing tours or hotels are created for search engines and not for customers. Sometimes texts full of keywords are simply impossible to read. No one reads such texts, but search engines see this and lower them in the search. That is, everything happens the other way around: excessive SEO optimization interferes with organic promotion. The golden rule is violated: texts should be for people, clients and tourists and not for search engines.
Help with education
This is partly the problem of modern education. If the education process is entrusted to robots, then this education will not teach students to think, and learning will take place according to a long-established pattern and there will be little benefit from such education. Moreover, many educational institutions are already expressing concern that students will soon stop writing essays and papers, and will copy them from the generative chat. But the problem is that any work must have at least novelty. Moreover, most likely, sooner or later, search engines and anti-plagiarism services will learn to recognize texts written by AI.
Help in writing articles
In this case, you will not envy the editors of scientific journals. If in the near future one does not learn to recognize texts written by AI, then the scientific world will be overwhelmed by a huge wave of large publications with zero scientific value, but well and competently written. There is no benefit from such articles for science, but compared to other works written by people, the article may become more popular due to the fact that it will formally look better. However, there is already a first reaction, for example, the World Association of Medical Editors (WAME) has issued special recommendations for the use of AI in writing scientific papers More.
Since the chatbot is not focused on indicating sources and is rather free with them, valuable information can be distorted, which is critical in engineering, medicine, and other precision industries. Therefore, restrictions are already being imposed, such as those introduced by the Stack Overflow knowledge exchange platform. More.
An example from practice
During a trip to Israel, the author got on an excursion to the Temple Mount. The guide, although he was a man, worked like a programmed robot. His task was to bring guests, tell the learned text and take them back. No information about the nuances of staying on the temple mount, where you can not go in shorts or with bare shoulders, he did not tell tourists. As a result, many simply could not get there and they had to urgently look for some kind of scarves, capes, etc. This is an example when a person behaves like a machine and presents the work of a guide as a chore. Naturally, such guides can be easily exchanged for robots, but this is not how we imagine a guide and we don’t want to visit such excursions … In any case, robots will not become more humane than people, so real professionals in the field of tourism are not in danger of losing their jobs.
Of course, progress cannot be stopped and language models will improve and perhaps someday they will reach a level where they can replace a person in routine work. However, tourism should get rid of this kind of work, because tourism is passion, inspiration, competition, challenge, transmission of culture and history. People do not go for standard sensations and impressions, they want to be given a human experience. And this should not be forgotten either when conducting excursions, or when compiling descriptions of tours and routes, or in any communication between the tourism business and its client, namely the tourist, since he / she is the main actor in the entire industry.
P.S. Large language models are like parrots that can speak human language, but it seems that no one has yet thought of talking about their business and attracting customers with the help of parrots. Although there is one example from the book “Treasure Island” by Lewis Stevenson, where the parrot knew only one, but very popular word: Piasters, piastres, piastres … We all hope that the tourism businesses will not look like such parrots.
Author: Dmitriy Tin – founder of Center Smart Tourism GmbH
Editing and translation into English: Leonid Andrianov