Innovation and AI: An Opportunity for Spanish Tourism in the Post COVID-19 Era

Abstract

The value of innovation in economic activity is undeniable, as is the importance of tourism in the Spanish economy. However, tourism has traditionally been considered less innovative than other sectors, which has led to less attention being paid to research in this area. With the arrival of the pandemic and its destructive impact on tourism, interest in new technologies applied to the sector has reached its peak. For this reason, this paper seeks to determine whether the use of these technologies, and specifically AI, represents an opportunity for the Spanish tourism sector in the post-COVID-19 crisis. To this end, a descriptive analysis of several successful cases of the application of this technology in the sector is carried out, as well as a solution to one of the problems arising from this new situation: the use of face masks.

1 Introduction

Innovation has proved to be one of the main drivers of economic development, which has led to a great deal of research in this area, in order to respond to changing market needs and trends. In recent years, technological advances have been made in various sectors of the economy. Among these, the “emergence” of Artificial Intelligence (hereinafter AI) stands out. Although its discovery dates back to the 1950s, it is only in recent decades that there has been a growing interest in this technology, due to the possibility of accessing data, the appearance of new and powerful hardware and software for generating algorithms, and so on. In this context, and despite the fact that the tourism sector is presented as a very traditional industry, it is one of the sectors with the greatest need and capacity for innovation. However, research on the technology in question has paid less attention to its application in the service sector than, for example, in the manufacturing industry. As a result, a large part of tourism research focuses on the study of other variables, such as the number of arrivals, employment in tourism or the seasonality that characterises this activity, which means that those studies that refer to the application of AI in the sector are more recent.

The arrival of COVID-19 plunged the economy and society into a period of great uncertainty. This uncertainty has had disastrous consequences for the tourism sector. The negative impact of the pandemic has also been exacerbated in countries that are highly dependent on tourism, such as Spain. According to the National Statistics Institute tourism ended 2019 with a contribution of 14.6% to Spain's GDP. As a result of the pandemic, this percentage will fall to 5.5% of GDP by 2020, representing an economic loss of more than ninety million euros (Instituto Nacional de Estadística, 2020).

This unexpected and unfortunate situation has necessitated the implementation of ongoing restrictions as a means of combating the virus. These include the compulsory use of masks, the limitation and control of capacity and the maintenance of safety distances. All of this, together with restrictions on the free movement of travellers, has had a significant impact on tourist activity. To this end, technology, and in particular AI, has helped to overcome many of these obstacles, leading to an acceleration of the innovation race in tourism.

2 Theorical Background

2.1 Innovation Trends

Towards service innovation: from the Oslo manual to the new wave of technology

The concept of innovation as a factor of development and economic growth has been studied over time and has been the subject of numerous economic theories. Until 2005, the most classic definitions of innovation, by authors such as Smith (1776), Schumpeter (1934), Drucker (1985) and Porter (1991), were developed with the industrial sector in mind and were used indifferently for the services sector, despite the specificities of this sector. With the publication of the third edition of the Oslo Manual this year, an important difference for our field of study was introduced: the inclusion of the service sector. This was not the case in previous editions. Two ideas emerge from this fact. The first is that studies on innovation in services are becoming increasingly interesting. Secondly, as the OECD notes, the lack of economic literature on innovation in tourism prevents the existence of a handbook, such as the Oslo Handbook, which collects comparable and generally accepted data on innovative activities in the sector (OECD and EUROSTAT, 2005).

The Spanish economy is currently showing a strong and growing commitment to new technologies. Within this framework, a new paradigm is emerging. These are tools that work with large volumes of data, which are ultimately transformed into highly valuable information for decision-making. Within these new trends, and given the recent interest it has generated in the tourism sector, this paper focuses on AI. Although it was born as a discipline of computer science, its study and application are generating a growing interest in the tourism sector.

2.2 AI in the Tourism Sector

Innovation and AI in tourism

There has often been a tendency to think that this sector requires little innovative activity. However, as the tertiarization of the economy progresses, this idea is beginning to be invalidated, confirming the need for the production and application of innovation in this sector (Mullo Romero et al., 2019). Thus, in 2011, tourism activity was declared as increasingly defined by innovation (UNWTO, 2011), defending a tourism strategy based on data. Likewise, an increasing number of experts consider AI as a fundamental part of strengthening Spain's leadership and competitiveness as a tourist destination, highlighting its strengths and trying to mitigate its weaknesses (Pedreño and Ramón, 2019).

An overview of the tourism sector

In order to understand the context in which the sector operates, it is worth recalling the specific characteristics of the tourism product, which are different from those of the industrial product and which partly explain the lower level of adoption of this technology compared to other sectors. The tourism product is intangible, heterogeneous, perishable, highly seasonal (Middleton and Clarke, 2001) and subject to constant change. It therefore not only requires innovation, but it does so in a dynamic, uncertain, complex and changing environment. It is not easy to operate in such an environment. That is why the tourism sector, in such a delicate period where the only certainty is uncertainty, needs tools such as AI to facilitate decision making.

Measuring AI in tourism

In the report "Realidad y perspectivas de la IA en España" by PwC and Microsoft (2018), the tourism sector is one of the twelve sectors in which AI will have the greatest impact. In this sense, it would be interesting to measure the real impact of this technology on the sector.

However, there are still no tourism-specific indicators to measure this [Aires and Varum, 2018]. Some experts in the sector have already warned that this will require "an evolution in digitalisation in order to work well in this new sector" (Mario Villar, 2021). In the meantime, various indicators and general reports can be used. For example, the "AI Index Report" (Stanford University, 2021) or the dossier "Indicadores de uso de Inteligencia Artificial en las empresas españolas" by the Spanish government, prepared with data from Eurostat, which shows that Spain's AI performance is below the global average, ranking 23rd out of 26 countries considered, although it has improved its position in several of the indicators included in the index, such as the rate of AI hiring, the number of LA companies created or the total private investment in this technology. At the EU level, "Indicators of the use of artificial intelligence in Spanish companies" is published, which concludes that Spain is among the EU-27 countries with the lowest level of adoption of AI in companies.

Evolution of interest and investment in AI for Spanish tourism

Despite the boom in interest and use of AI in Spain, its performance is still below average. In 2020, the Secretary of State for Digitalisation and Artificial Intelligence promoted the National Strategy for Artificial Intelligence (ENIA), with an investment of more than 600 million euros for the period 2021–2023. In parallel, some autonomous communities-such as the Valencian Community, Galicia or the Basque Country are presenting their own AI strategy (Gobierno de España, 2020).

In the field of tourism, the Secretary of State for Tourism is promoting the network of smart tourist destinations through SEGITTUR (2021). Likewise, several municipalities are already launching their own initiatives, such as the Digital District of the Community, which is leading conferences on innovative solutions for tourism (Business Insider, 2021).

A stumbling block: AI ethics and regulation

Many experts agree that the Spanish tourism sector has what it takes to continue growing with AI. "We have the opportunity to lead artificial intelligence in Europe from Spain”, so says Reyes Maroto, Minister of Industry, Trade and Tourism of the Spanish government. However, it has been proven that Spain is currently far from being considered a leader in the application of this technology. The idea of global leadership, with competitors such as the United States and China, may be too ambitious. One of the obstacles is "ethics", and when it comes to personal data, this becomes an essential factor to be taken into account. In addition to the White Paper on AI, the European Commission (2021) has recently approved the first regulation on AI. In the specific case of Spain, the General Law on Data Protection (LGPD) has been added. In some autonomous communities, such as the Valencian Community, there is even a specific regulation.

3 Aims and Hypotheses

The main objective of this paper is to study the role of innovation, and specifically AI, in the post-pandemic future of the Spanish tourism sector. In this sense, a series of specific objectives have been set. Firstly, to study the evolution of the concept of innovation in the tourism sector. Secondly, to identify the new trends that are currently shaping the tourism sector. In addition, to identify a series of cases in which AI offers a solution to the problems caused by the pandemic in the tourism sector. Finally, to propose its own solution to one of the problems arising from this new situation, the use of masks. Based on these objectives, the following hypothesis is formulated Does the use of innovation, and specifically AI, represent an opportunity for the Spanish tourism sector in the post-COVID-19 era?

4 Methodology

This study presents an analysis of the opportunities introduced by innovation, especially AI, to the Spanish tourism industry since the onset of the pandemic and the potential opportunities for the future. The study is divided into three stages.

First stage: Collection of AI application cases as a solution for post-COVID-19 tourism. A search for AI-based technological solutions for tourism recovery was carried out in various written sources, all of which are cited in the bibliography of the paper. We also attended the masterclass "IA en el día a día", organised by Spain AI, and several webinars, including "VIII Thinktur Technology Transfer", by Thinktur, and "Aplicación de la Inteligencia Artificial al turismo tras el COVID-19", organised by Turisme Comunitat Valenciana.

Second stage: A selection and study of successful cases was conducted for each of the five segments, building on an accurate classification of typologies of AI solutions for the sector as defined by Thinktur and also including an additional segment for sustainability.

Third stage: A proprietary solution was proposed using Teachable Machine. Finally, a proprietary solution has been suggested utilising Teachable Machine–an AI tool crafted by Google which simplifies the creation of models- in response to a recurring predicament in this contemporary era: regulating the correct usage of masks within enclosed surroundings.

5 Results. Case Studies and Proposals

5.1 Tourism and COVID-19

In 2019, Spain reclaimed its position as the top global leader in tourism competitiveness, as reported by the World Economic Forum (2019). Regrettably, the high hopes for growth in the sector for the year 2020 were hindered by the emergence of a new virus. Aena's (2020) revealed a sharp decrease in air transport passengers from 275,247,387 to 76,064,322, representing a staggering decline of 72.4%. Undoubtedly, the pandemic has had a significant impact on bars, restaurants and other catering establishments, resulting in intermittent activity interruptions due to restrictions on opening hours and capacity. These restrictions have forced such businesses to temporarily cease their activities and, in the worst cases, to shut down. The Bank of Spain has reported that 50,000 tourism businesses were closed in 2020 due to the pandemic (Ministerio de Trabajo y Economía Social, 2021). The state of the tourism industry has raised concerns about the effectiveness of the Spanish tourism model. A key priority now is to modernise and revamp the sector. Accordingly, the Spanish Government (2021) has emphasised the crucial requirement for "a modernisation and enhanced competitiveness strategy to prepare the industry for significant changes, particularly in the areas of digitalisation and sustainability." There is an opportunity to promote a data-driven strategy, in which innovation, especially AI, is crucial.

5.2 Cases of Application of AI in Tourism COVID-19

In the arena of industry resizing, solutions like BiOnTrend have surfaced. BiOnTrend is a collaborative data analytics tool for hotels designed to address the sector's dearth of a proprietary tool that is competent in scrutinizing authentic data on tourist accommodation and destinations (HOSBEC and Turisme Comunitat Valenciana, 2020). Tools have been developed to ensure compliance with minimum interpersonal distancing in real time (Landing AI, 2020). Additionally, measurement and control of beach capacities has been achieved through the use of technology (Alicante Plaza, 2020; Konica Minolta, 2020).

For communication and transparency, Intelligent Virtual Assistants (AVI) or chatbots are particularly noteworthy. A noteworthy accomplishment is the Carina chatbot, created by the technology firm 1MillionBot based in Alicante. The chatbot can respond to inquiries about over 300 subjects regarding the COVID-19 outbreak with a success rate of 91.70% (1MillionBot, 2020). Concerning health, thermographic cameras permit swift detection of fever without direct contact with the person (Tecon Group, 2020). Finally, in regard to sustainability, Substrate AI and the Poseidon hotel chain have initiated a project with the aim of reducing their hotel energy consumption by 10% through the application of AI (Hosteltur, 2021).

5.3 An Own Case Study: Mask Control with Teachable Machine.

Several companies are dedicated to developing advanced technologies and making them widely accessible to the public. Google, a technology leader, offers the Teachable Machine tool that enables the creation of automatic learning models in a straightforward, user-friendly manner. In the following section, we will showcase an applied case study that addresses one of the most significant changes resulting from the pandemic: the mandatory use of face masks. The proposed AI-based solution is capable of detecting correct and incorrect mask usage.

Upon accessing the tool, users must select one of three project types: audio, posture or image (in this instance, image has been selected). The model follows a five-phase structure that includes learning, preparation (or training), evaluation, parameter reconfiguration and model export.

Phase (1) Learning. Two classes of data will be created, with different labels, namely "mask" and "no mask." Subsequently, pictures will be entered via webcam. The initial phase involves gathering data, which is crucial to a large degree for the model's success. The device will learn from the samples provided to it. The user will capture a collection of photographs depicting themselves wearing a mask. The first sample consists of 31 images in total.

To obtain the "no mask" class, the previous procedure will be duplicated, but without utilizing the mask accessory shown in the other class. As a result, the ensuing sample consists of 27 photos. By doing so, the primary fundamental data will have been provided to initiate model learning. It is not desirable to consider practices such as wearing a mask without covering the nose, mouth or holding it on the chin as acceptable. As a result, a new collection of images will be included in the "no mask" set, demonstrated as follows.

Now that the datasets have been defined and supplied, the training or model preparation phase will commence.

Phase (2) Preparation or Training. Once the "prepare mode-lo" option is selected, the mode-lo will commence processing the data. Essentially, the program will amalgamate all the given data to generate algorithms, which subsequently can be employed to automatically detect if a person is wearing a mask appropriately or not from an image. Following the completion of the model, a "Preview" block will manifest on the right-hand side of the page. This is the result of the model's evaluation phase.

Phase (3) Evaluation. If everything has functioned properly, it should be possible for the model to identify, from an image, whether or not the individual in the picture is equipped with a mask. Figure 1 demonstrates the model's accurate identification of the case in question. Additionally, it effectively discerns instances where the mask is not securely fastened.

Fig. 1

Source Screenshot from the Teachable Machine website, 2022

Correct detection of mask use/non-use.

However, the model may also present faults. If the user swaps their surgical mask for an FFP2 mask, the model fails to detect accurately. The same issue may arise if a woman wears Niqab as the model detects the presence of a mask, even when there isn't one. This occurs due to the fact that we haven't educated the model on the difference Here is a further indication of the significance of the preparatory stage for the model to acquire and develop knowledge to the fullest extent feasible.

Phase (4) Parameters Reconfiguration. The issues identified during the evaluation phase will be resolved in this stage. The model will be continually fed to fine-tune the results. As the model is currently unable to accurately recognise the FFP2 mask, additional images will be added to the "mask" label, featuring users wearing FFP2 masks. After this step, the model will be re-run, and the updated results will be examined. Figure 2 depicts that the FFP2 mask is now successfully detected.

Fig. 2

Source Screenshot from the Teachable Machine website, 2022

Failure of the model when using another mask or Niqab, and new model corrections.

Phase (5) Exporting the model. Teachable Machine provides the option to export the model for personal use. Once the model is functioning correctly, it can be downloaded, uploaded to Google Drive, shared via link, and even copied in JavaScript format for application in a web environment. It is worth noting that this proposal represents a simplified version of a machine learning model based on AI. According to current regulations on Artificial Intelligence in Europe, the proposed solution would only be viable for use in private spaces, such as hotels, restaurants, shops, and museums, due to ethical considerations. The bibliographical review has highlighted the prohibition of image recording in public areas under the same legislation.

6 Conclusions

All in all, it is concluded that there are reasons to believe that AI represents an opportunity for the country's post-COVID-19 tourism. However, in order for the sector to benefit from these opportunities, a number of challenges need to be addressed, which are detailed below.

• AI measurement system for tourism. Regardless of the variable under study, measuring and evaluating the results is an essential step in order to move forward. At present, AI does not have a measurement system that makes use of tourism-only indicators, which makes it difficult to establish corrective actions that seek to improve the sector's results.

• AI ethical framework for tourism. The lack (or sometimes over-regulation) of AI regulation can be a disadvantage in the race for innovation. Although the European Union is making progress in this area, work must continue on the construction of an ethical framework that adapts to the specific needs of this sector, avoiding leaving behind countries with great tourism potential, such as Spain.

• AI Education and Training. There are individuals in our society who are hesitant to implement this new technology, likely due to fear or lack of knowledge. Therefore, it is critical to educate and provide training on AI to the population, allowing them to comprehend the benefits it offers. By doing so, utilising AI in the tourism industry could lead to the development of higher quality employment, a reduced workload, and overall enhance the service provided to visitors. A clear example is the case study proposed in the results section.

Is it wise to invest in a technology that lacks public trust? This is a valid question. To mitigate this issue, there should be a focus on creating new job opportunities and facilitating reskilling for those who lack technological proficiency.