Strategic Assessment

Keywords: ChatGPT, public diplomacy, hasbara, artificial intelligence (AI), disinformation, Israel

Introduction

The presentation of the ChatGPT chatbot by OpenAI Ltd. in November 2022 aroused huge interest worldwide because of the application’s advanced capabilities with the potential to improve the quality of life (for example, by increasing the precision of medical diagnoses) or, conversely, to be used for malign purposes (for example, to conduct more effective phishing attacks with greater ease). The development also has considerable potential significance in the field of international relations, and particularly public diplomacy, since it underscores the ever-increasing relevance of cyberspace for relations between countries, peoples, and individuals—a trend evident for the past two decades. The use of artificial intelligence (AI) to produce deepfake video clips, for example, during the war in Ukraine, could affect public perceptions of events in the war or the overall situation.

Since its establishment, the State of Israel has faced the ongoing challenge of managing its image in the eyes of the world, and over the years has invested considerable efforts and resources in public diplomacy. Nevertheless, many among the Israeli public sense its performance in the perception management field to be sorely lacking, since there are many anti-Israel organizations throughout the world that try to undermine its legitimacy as an independent state.

The use of AI for public diplomacy involves both opportunities and risks. Technological advances on the one hand, together with Israeli society’s concern for its public image, create an opportunity for public diplomacy to promote the use of AI as a means of improving its performance. Improvement could be manifested on three levels. On the strategic level, the use of AI can enable Israel to gain a better insight into possible future global changes that could change the balance of world power and affect how Jerusalem might manage its foreign relations, particularly with the great powers. On the intermediate level, AI could provide a response to the existing bureaucratic maze of the public diplomacy system, which often prevents Israeli influence campaigns from operating effectively. On the tactical level, AI enables the elements engaged in providing information to respond in a faster, better way to attempts by social media anti-Israel influencing campaigns at disinformation and delegitimization, including with actions that are themselves based on artificial intelligence. The article ends with a number of recommendations that could start the process of AI’s absorption into the campaign.

The Strategic Level: Deepened Understanding of Demographic Changes and their Political Significance

Traditional diplomacy is based principally on the activities of official state representatives, but the internet and social media have given non-governmental elements, business organizations, and social groups much more scope to shape international relations. For that reason, social and demographic changes within populations may have a more significant impact than before on the objectives of public diplomacy, joining the classic considerations of the balance of power between states and their national interests.

Demographic changes in a country can be reflected by a growing or shrinking population (for example, the populations of China and Europe, both currently considered economic powers, are expected to drop by the year 2100), or by a change in local population composition (migrants comprise an increasingly large proportion of the populations of Europe and the United States). Changes in population size can have a direct effect on the components of a country’s hard power (for example, more or fewer people of working age), while changes in population composition can lead to policy changes due to changes in voting patterns that result in elected governments with different approaches to foreign affairs.

Demographic changes have special significance for the public diplomacy efforts of small countries such as Israel. These countries apply two main means of maintaining their power in the international arena: persuading large powers to work on their behalf to help them realize their national interests; and creating a coalition of several small states to promote common interests. Official and unofficial entities hoping to achieve such objectives must on the one hand establish and maintain close ties with official decision makers, and on the other hand, enlist the support of physical and online communities.

Naturally, demographic changes oblige these entities to become more familiar with the nature and effect of the changes. While diplomatic personnel in Israel have identified the need to prepare for global demographic changes, this does not appear to have encouraged changes in the definition of target audiences. This can be seen first and foremost in the United States, where Israeli public diplomacy prefers to address groups traditionally identified as pro-Israel (Evangelicals and Jews) rather than young African Americans and American Hispanics. And in any case, the situation is complex. Studies from recent years show that in addition to the high proportion of African Americans and Hispanics who hold antisemitic views, there is also considerable support and sympathy for Israel and the Jews. At the demographic level, however, while the proportion of Evangelicals in the US population fell between 2007 and 2021, the Hispanic population is the fastest growing community in the country.Nevertheless, if there are any plans to address the Hispanic community, the activities are generally localized or abandoned due to lack of resources.

At the strategic level, it is possible to use ChatGPT to produce a range of demographic forecasts and in-depth analyses of their political significance. In the study of climate change, for example, ChatGPT is identified as a means for promoting understanding of the nature of changes and improving the accuracy of climate forecasts, using data from many different sources. Similarly, it is possible to use ChatGPT to create highly accurate scenarios to describe the political consequences of predicted demographic changes. The creation of such scenarios requires participants to demonstrate skills of planning and execution, and these skills are already recognized as a possible tool for diplomats in training.

It is also possible to use ChatGPT to gain deeper knowledge of the foundational ideologies and values of selected target communities. In the US, for example, the views of non-white populations differ from those of the white population, which has been a more familiar audience for Israel over the years. Failure to recognize these differences could lead to erroneous perceptions about the various communities when planning and executing Israeli information campaigns, thereby damaging their effect.

Deeper understanding of target communities can be achieved by using ChatGPT for sentiment analysis and stance detection. Sentiment analysis is the use of AI-based language processing techniques to mine and analyze texts in order to identify and quantify the emotional conditions they express. Stance detection is considered a more advanced capability, since it requires characterization not only of the emotions expressed in a text but also of the writer’s position on a specific topic. Therefore stance detection is deemed a vital tool for the analysis of public opinion on social networks, particularly on political and social issues. In 2023 studies were published that compared the performance of ChatGPT in these tasks with other language models, showing that ChatGPT is capable of more accurate sentiment analysis. As for stance detection, ChatGPT is considered the best model currently available, which could even change the research paradigm in this field.

At the strategic level, therefore, it can be argued that the use of artificial intelligence could enable every country to predict demographic changes, prepare for their political consequences more accurately, and define optimal principles for spreading information and taking political action. These benefits are particularly important for small countries like Israel: because of their dependence on the great powers to realize their interests, they are strongly affected by changes in the global balance of power, in which demographic changes play a central role. At the same time, understanding the political importance of changes could affect decisions by small countries over whether to work toward new alliances with other countries, or to shift the nature of existing international coalitions of which they are members. In Europe, for example, the populations of countries such as Latvia, Lithuania, Hungary, and Poland—which in recent years were a preferred political target for Israel—are likely to decline, while the populations of countries such as Sweden and Ireland, which in recent decades experienced diplomatic crises with Israel, are expected to continue growing. This could encourage Israel to aim for stronger links with demographically growing countries, at the expense of or at least alongside promotion of the alliances with Eastern Europe.

The Intermediate Level: Improved Processes for Planning and Execution in Public Diplomacy

Optimum understanding of international strategic changes and policies formulated to support public diplomacy, as good as they may be, cannot help a state’s performance on the global stage if the people responsible for execution encounter bureaucratic difficulties at a practical level.

The extensive efforts that Israel invests in public diplomacy include the use of new technologies: the Foreign Ministry is promoting a wider use of a range of digital capabilities, for example by including intensive training in the field of digital diplomacy in its training course—making it one of the most influential Foreign Ministries in the digital realm.

Nevertheless, Israel’s ability to achieve success in this field suffers from bureaucratic challenges. Israeli public diplomacy is a multi-pronged system comprising numerous government entities, which over the years have acquired powers at the expense of the Foreign Ministry. Although the National Public Diplomacy Directorate was established to coordinate Israel’s public diplomacy activity, it is hard pressed to fulfill its mandate, and in effect there is no central management basis for the activity that can define its strategic objectives. There are also planning problems when government ministries work with external organizations. One example was a venture called Building Permit set up to reinforce the dissemination of pro-Israel information in US academic institutions, under the direction of the Foreign Ministry and an external association, with no examination of its achievements, and in spite of an internal assessment that its benefits did not justify its high cost.

The use of ChatGPT could help Israel improve the processes of planning, implementation, and real-time update of influence campaigns derived from its strategic planning. This could be done, for example, by providing recommendations on the optimum division of tasks among the project team and presenting assessments regarding changes that must be taken into account (such as possible delays in project completion). ChatGPT can also be used to collect and summarize data from multiple sources, particularly on matters relating to improved project management, such as the various methodologies available. And finally, ChatGPT can be used to set up a central communications platform for everyone involved in public diplomacy. Such a platform could not only limit cases of misunderstandings, based on one participant receiving partial information, but also create transparency in decision making processes, and thereby strengthen the trust of the public and participants in public diplomacy functioning.

The Tactical Level: Improved Ways to Counter Anti-Israel Influence Campaigns

In addition to the possible benefits at the strategic and intermediate levels, there are also ramifications of using artificial intelligence for specific influence campaigns: although it can intensify the threat of false information being disseminated on the internet, it can also provide enhanced solutions for dealing with the problem.

AI can upgrade the creation and distribution of false content on social media in two important ways. First, it can extend the reach of false information publicized, by increasing the quantity of content posted and by increasing the number of hostile elements able to do so. While the spread of false information without the use of AI is usually based on publishing the content on several platforms simultaneously—a repetitive process that can create burnout and cognitive overload for creators, ChatGPT enables malign elements to create multiple versions of content for publication on multiple platforms. At the same time, ChatGPT developments can enable malign elements to significantly extend the scope of potential topics for false content.

Moreover, it is possible to increase the scope of elements engaged in spreading false information, because the adoption of AI tools reduces the cost of advanced means for creating and distributing the material. Without AI, the cost of implementing a large-scale influence campaign on social media is estimated in the tens of millions of dollars, making it beyond the reach of many malign developers. With AI, the cost of creating high quality fake pictures and audio or the cost of cross-platform testing to test the reaction of social media users to an influence campaign is far less. Even before ChatGPT appeared, it was known that anti-Israel elements had the ability to join together to spread their messages widely, becoming very prominent on the web. Lowering the costs of advanced influence campaigns could therefore widen the quantitative gap in their favor and reinforce their social media presence, to the detriment of elements working for Israel’s benefit.

Second, the use of AI allows hostile parties to improve their disinformation content, making it more convincing to target audiences. In particular, ChatGPT-4 output is considered even more convincing and detailed than version 3.5. But the improved quality of the false content not only raises the chances of influencing selected target audiences, but makes it harder to locate and remove. ChatGPT can also be used to plant disinformation about historical events, by forging and distorting historical documents. This ability has special significance for Israel, which uses memories of the Holocaust and the historic right of the Jewish people to its country as central messages in its public diplomacy.

The significance of these developments is that a wider range of hostile elements can spread greater quantities of better quality false information on social media, particularly relating to the central messages of Israeli public diplomacy, with less chance of discovery by the social networks, enforcement agencies, and users.

Tackling the distribution of ChatGPT-based disinformation has become a widespread global activity in recent years, focused on AI-based influence campaigns. Among the variety of initiatives and tools used in this field it is possible to identify three possible channels of activity, by means of which public diplomacy can recruit ChatGPT to provide a better response to disinformation campaigns, including those based on artificial intelligence.

The first channel is the development of technological means for identifying ChatGPT-based content, such as GPTZero: an AI-based tool that uses the ChatGPT model to identify whether a particular text has been written by a person or by AI. While GPTZero does achieve high rates of success in characterizing long texts, it has difficulty analyzing short texts such as tweets on X (Twitter).

The second channel is the use of ChatGPT for the fast creation and spread of public diplomacy content: posts on social media, press releases, and even official speeches. The Foreign Ministry first used artificial intelligence for public diplomacy purposes in January 2023, when it was assisted by a platform developed by Israeli start-up D-ID to produce short videos in eight languages, including Persian, Chinese, and Arabic.

In some cases, the quality of messages generated by ChatGPT is very similar to that of messages produced by human lobbyists. In fact, there is concern that the process of using ChatGPT to generate content could be done without proper attention to the formulation of the product. In the world of diplomacy, in which every nuance has political significance, the selection of inaccurate terminology could lead to gaps in expectations and in communication between countries, even to a diplomatic crisis. At the same time, inattention to text composition could mean that certain content is perceived by its target audience as too cliched or inauthentic. In the case of Israel, use of ChatGPT could lead Israeli elements to assimilate terms that appear neutral but in fact serve the purposes of anti-Israel entities. An example is the adoption of the word “movement” by the BDS organization, which gives it the image of a legitimate social activity in the eyes of foreign audiences. The more such terms appear in ChatGPT-generated content, the more Israeli public diplomacy may unconsciously promote anti-Israel narratives.

This issue is linked to a broader problem—the problem of inherent bias in large language models. These models, including ChatGPT, are trained to analyze large quantities of texts and use them to generate new material. In the training process, developers feed the models with huge amounts of data from a variety of sources, such as books and internet sites. Sometimes the resulting content generated by the models reflects offensive or unfounded views based on gender, race, or social group, so that the use of AI leads to biased decisions that perpetuate prejudices. In the field of foreign policy, ChatGPT could bring together messages that are offensive to certain target audiences, which would of course damage Israel’s image. It is therefore advisable to use ChatGPT to generate rapid drafts of messages with the final polish done by human beings.

The third channel of activity is the use of ChatGPT for fact-checking of content for publication in the media. Like the two previous channels, here too the findings on the performance of ChatGPT are mixed: while in most cases ChatGPT is able to correctly classify statements as true or false, its performance is not consistent, so that a specific claim that is checked twice or expressed in two different ways could be rated differently. Accordingly, fact checking organizations claim the main advantage of using AI in their work is the ability to identify more claims that require checking. This need has become more pressing not only because the use of AI can increase the reach of disinformation, but also because the resources available to fact-checkers are considerably less than the resources of the producers of disinformation. Fast refutation of lies with facts can help Israel to limit their spread, and thus minimize the possible damage to Israel’s image caused by various events.

A noteworthy Israeli technological development on this subject is SAVEE, which enables users to respond to Holocaust denial content on social media using generative AI to compose the response. Fact-checking initiatives like this can produce not only rapid reactions to anti-Israel and antisemitic messages, but also give any interested individual the power to check claims from malign elements, without special training. This will considerably increase the fact-checking potential of the public diplomacy system on social media—a need that was identified as essential following Operation Guardian of the Walls (May 2021).

For defined influence tasks, public diplomacy professionals can derive particular benefit from using ChatGPT for fast and effective reaction to specific events (such as a military operation). Speed is a vital component when refuting false information on social media, and the rapid production of public diplomacy messages together with optimum allocation of the necessary auxiliary tasks helps to ensure it is achieved. It is also possible to recruit ordinary citizens quickly who are not usually part of the public diplomacy system, using ventures similar to SAVEE. The improved speed of response thanks to all these components will ultimately help in tackling the problem of the quantitative gap in favor of anti-Israel elements.

Policy Recommendations

The strategic and practical benefits of artificial intelligence can serve anti-Israel elements just as they can support Israel. Government and public diplomacy people are therefore recommended to act quickly to assimilate the use of AI in the public diplomacy system, taking account of the challenges involved. The following four courses of action could be part of an initial plan:

• Government support for the development of enhancements to ChatGPT capabilities: Existing ChatGPT capabilities can improve performance in tasks focused on influence, but there are important potential risks. The public sector is therefore advised to implement schemes that will encourage technology companies to seek effective and innovative responses to the problems. This could be done via hackathons (intensive entrepreneurial events where participants develop technological solutions for challenges defined by the hosts), such as the one organized by the Canadian government in 2023. The publication of government public appeals for research in the field of AI, such as one from August 2022, could also contribute to this solution. Future appeals should focus on research into the particular advantages and drawbacks of ChatGPT that are relevant to public diplomacy. This research could lead to the development of numerous enhancements, thanks to the large number of Israeli start-ups in the field of generative AI.
• Definition of guiding principles for controlled use of ChatGPT: Because of existing problems of imprecision and bias in AI, government ministries should define principles to ensure full utilization of the potential benefits, while limiting the occurrence and impact of problems. For example, it is possible to stipulate that any chatbots used must meet the criteria included in the European Union AI Act, currently in the process of ratification by EU institutions. According to this law, companies developing chatbots, such as OpenAI and Google, must comply with a series of strict regulatory requirements, such as ensuring the ability of human operators to override how they work before they are marketed commercially.
• Response to the bureaucratic challenges of inter-organizational use of AI: Public diplomacy involves numerous elements, mainly in the public sector. An existing study shows that the adoption of AI for inter-departmental collaborations in the public sector is likely to encounter problems relating to the necessity of data sharing (for example, due to differences in privacy policies). Such problems may be more pronounced when government ministries work with non-governmental organizations, due to gaps in the governing procedures and laws pertaining to the various parties sharing data. It is therefore recommended that AI be used to meet these challenges, for example through accurate definition of the problems to be resolved, in order to improve understanding of the types of information to be shared between organizations.
• Trial assimilation of ChatGPT in the regular work of the public diplomacy system: After defining the criteria for using AI in a government framework and overcoming the coordination challenges, it will be possible to use ChatGPT for limited experimental projects designed to improve distinct aspects of public diplomacy work. For example, it is possible to analyze the views of defined target audiences in a specific country over a limited period and examine the quality of AI performance. At the same time, it is possible to integrate the use of AI-based scenarios as part of Foreign Ministry training and compare the performance of trainees who experience them with that of trainees who have not worked with these scenarios.