Cdn.ymaws.com



Lessons from the Newsroom: A Case for the AI-Augmented Joint Information Center Deborah Grigsby SmithCentennial Airport (KAPA)Arapahoe County Public Airport AuthorityAugust 29, 2019AbstractMultiple reporters work on independent stories as well as collectively on enterprise or investigative pieces. Shorter news cycles, staff cuts, and the overwhelming inflow of information, news tips, fake news, social media, and breaking news mirror the battle rhythm in most joint information centers. Both must collect, process, vet, repackage, and distribute information, data, and imagery in rapidly evolving situations. There is one difference, and this is the fact newsrooms operate collectively with a centralized content management system that fosters a coordinated, sleek flow of information from the creator to published product. AI-based systems, fueled by the growing amounts of data now available, are rapidly becoming competent, autonomous public communicators, able to manage and distribute large amounts of information with accuracy and ease. As the stakeholder interface with technology deepens and machine-authored content becomes more prevalent, public information officers must begin to evaluate how AI and big data will reshape the profession. The challenge for future public information officers lies not in the ability to communicate, but in keeping up the accelerating technology used by the populations and stakeholders they serve. This study examines three case studies of major global newsrooms that employed artificial intelligence-based content creation and management tools to improve coverage, increase efficiency, and target specific audiences. It also presents the findings of a survey of 33 public information and communications professionals and their thoughts on the possible use of such tools in a joint information center environment. Keywords: artificial intelligence, machine learning, natural language generation, data, big data, public information, public information officer, joint information system, content managementTABLE OF CONTENTSIntroductionPage 4Problem StatementPage 5Research QuestionPage 6Literature ReviewPage 8Method: Case Study and SurveyPage 15ResultsPage 22Conclusion and RecommendationsPage 33Recommendation 1:Page 33Recommendation 2:Page 33Recommendation 3:Page 34DiscussionPage 34ConclusionPage 40ReferencesPage 41Appendix A: Primary Survey InstrumentPage 43Appendix B: Raw Data from Primary SurveyPage 47Appendix C: Secondary Survey InstrumentPage 55Appendix D: Raw Data from Secondary SurveyPage 56INTRODUCTION“The only sustainable response to technological disruption is to try to lead it.” —Enrique Dans CITATION Dan19 \l 1033 (Dans, 2019)A Case for the AI-Augmented Joint Information Center In times of disaster, timely and credible public information messaging can save lives and protect property. Generous and continuous access to data is essential for public information officers to not only understand a rapidly evolving situation but also to plan and develop the strategic messaging that supports emergency response. As human interface with smart technology deepens, and the use of social media and computer generated-content becomes more prevalent, public information officers and the joint information centers (JICs) they serve can quickly become overwhelmed parsing high-velocity mountains of information from informal, official and unofficial sources. This surge, especially in an already understaffed JIC, increases the probability of mistakes, ineffective, late, overlooked, or misaligned messaging. Artificial Intelligence (AI) with its ability to quickly process large amounts of data, facilitate human decision-making and even predict human behavior, have already made themselves at home in almost every aspect of life, to include emergency management, so why has it not found its way to the JIC? There is a growing number of case studies derived from the news industry, as well as the marketing and public relations field, that point to promising applications for its use in the joint information environment. The challenge for future public information officers rests not in the ability to communicate, but rather, in understanding, leveraging, and keeping pace with exponential amounts of data driving the technology used by the stakeholders they serve. This thesis explores how three major news agencies have employed AI technology to expand coverage, increase accuracy, and better connect with audiences, and how joint information centers could benefit from adopting similar technology.This intended audience for this document is the working emergency management and public information officer community. The researcher presumes the audience has advanced knowledge of joint information center job assignments, organization, operation, and challenges; hence there has been no attempt to provide introductory or explanatory material on the subject for casual readers. Additionally, discussions around artificial intelligence, within this document, will be kept at the conceptual level and not delve into specific technical, software, or hardware requirements. Problem StatementExamples of news-writing algorithms began to make headlines in 2014 when digital news editor and computer programmer Ken Schwencke developed a few lines of code that pulled alerts from the U.S. Geological Survey about earthquakes that exceeded a specified magnitude threshold. The algorithm, called Quakebot, would then extract the relevant data and plug it into a pre-written story template, much like a form letter. On the morning of March 17, 2014, Schwencke was rattled from his slumber by an early morning quake. Instead of diving for cover, he rushed to his laptop where he found a news story about the 4.7 earthquake, already written and waiting in the queue. He quickly proofed the text and hit "publish." Within three minutes, the story appeared online, making the Los Angeles Times the first media outlet to report the Westwood, California earthquake. It would also make the Los Angeles Times the first major media outlet to publish a news story written entirely—and autonomously—by a “robot reporter.” CITATION Ore14 \l 1033 (Oremus, 2014). Since then, more prominent news agencies have also employed similar AI-based technology to create content, post to social media, enhance customer engagement, increase revenue by customizing content to consumer preferences, and by driving article and advertising clicks. According to research studies, AI-based systems offer many promising applications for the news and other public communications field, automating tasks and augmenting human capacity, in particular. Information parsing and content creation are essential aspects of the public information officer’s job. Most spend a significant amount of the operational period gathering, writing, and repackaging information across multiple platforms to targeted stakeholder audiences. AI-based content creation systems or “robot reporters” as they are, sometimes, called are becoming increasingly competent, autonomous publishers and public communicators, capable of plowing through mounds of data and authoring stories with increasing accuracy and speed. They can quickly create infographics, short video clips, add captions, identify faces in images and quickly translate between languages—to including many regional dialects. While stakeholder demand for a steady feed of current information increases—particularly during emergencies—public information officers working large-scale disasters, both in their home agencies and in a joint information center, risk data saturation and must triage certain types of stakeholder engagements over others. Public information officers become relegated to keeping up with essential tasks, keeping them from more strategic and high-value activities. This practice can compromise the quality of service to the public.This thesis explores and offers a case for the use of AI-based technology in the joint information center environment, specifically automated content creation and management tools, similar to those used by newsrooms. By comparing the functions of a large newsroom with that of a large working joint information center, the benefits of AI augmentation to improve accuracy, drive productivity and enhance public messaging become more evident. In addition to everyday social media monitoring, these systems, when integrated with credible data sources, tirelessly mine data, schedule appointments, create reports, press releases, develop social media content, blogs, video, and other public information products autonomously. When networked, these systems may also help structure JIC public information products that are supported, and in step with, the individual agency or organization—and the emergency operations center. Using a small, well-qualified population, this research will also gauge how public information officers feel with regards to working with such technology. Research QuestionWhile a large segment of current AI research falls into the military, scientific, media, finance, and marketing category, the use of AI in the public information has received little scholarly attention. As marketing firms, national and international newsrooms—as well as nefarious and disruptive entities—grow more adept at using AI to engage audiences and influence human behavior, its potential for application in the public information field remains woefully untapped. This thesis presents the following research question: How can AI-based content creation tools used in newsrooms can be adapted to improve general information messaging and joint information center operation? This thesis does not purport AI or AI-based systems as a single, emergency-ready solution. It does, however, offer foresight of its potential to improve joint information center performance.Literature ReviewArtificial intelligence and AI-based algorithms have long been used by marketing research firms to gauge, anticipate, target, and influence audience behavior. AI has transformed, in just a few short years, the way information is collected, categorized, integrated, and redistributed into new products, services, and tools that support critical business and decision-making functions. (Kazuo Yano, 2017). As newsroom financial challenges force deep staff cuts and reorganization, AI technology and “robot writers” have stepped in to fill the gap. However, when it comes to AI, government entities, at all levels, often lag behind private sector successes.Federal Chief Information Officer Suzette Kent recalled a time when the government was a “leader in technological innovation.” She suggests federal agencies have fallen so far behind that they will have to work aggressively just to catch up with basic private sector practices. CITATION Bur18 \l 1033 (Burr, 2018) Kent states, “agencies will have to pursue decades-long data and IT modernization plans so that government services meet the expectations that citizens have cultivated from commercial tech experiences.” CITATION Bur18 \l 1033 (Burr, 2018) Failure to keep up with technology, according to Sam Blakeslee, head of the Cal Poly Institute for Advanced Technology and Public Policy, can result in a troubling disconnect between government and its citizens. Referring to the ubiquity of technology, such as smartphones, Blakeslee points out, “this technology revolution goes far beyond the amazing new gadgets we wear, carry, and touch almost every hour of every day. The revolution didn’t stop at giving us neat new George Jetson-type devices. It went quickly toward deconstructing the entire economic ecosystem that we once knew as our status quo.” CITATION Eid15 \l 1033 (Eidam, 2015)Artificial Intelligence a National PriorityDriven by large amounts of available data and increased computing horsepower, AI has developed wide-reaching applications from marketing, to finance, news, transportation, critical infrastructure, science, health care, and the military—all of which drive the American economy. In February of 2019, President Donald J. Trump signed an executive order that laid out a broad plan for American dominance in AI. The presidential directive, which comes at a time when relations between China and the U.S. are strained, does not allocate any funding but does call on federal agencies to prioritize their existing funding for AI projects. In 2016, the Obama administration released a report on the future of artificial intelligence and a strategic plan for federally-funded AI research, titled “The National Artificial Intelligence Research and Development Strategic Plan.” The National Science Foundation announced on June 21, 2019, that it would join federal partners in reevaluating the plan’s priorities to add a partnership building element. According to Jim Kurose, NSF assistant director for Computer and Information Science and Engineering and the co-chair of the AI Subcommittee, the AI field is interdisciplinary and cross-sector by nature. (NSF, 2019) Kurose asserts public and private partnerships provide much-needed opportunities for organizations to leverage the flow of people and knowledge among academia, industry, and government to enhance economic growth and global competitiveness as well as emphasizes the value of expanding research and education in AI. (NSF, 2019)Artificial intelligence (AI) is, according to former FEMA Deputy Administrator Richard Serino, perhaps the most powerful of emerging new technologies and it is increasingly being put to use in the service of emergency management. CITATION Ser18 \l 1033 (Serino, 2018) He states, “in fact, there has been such a profusion of disaster-related solutions imbued with various artificial intelligence and data capabilities that emergency managers are under increasing obligation to develop the skills to effectively assess the relative merits of various technologies.” Private Sector Investment in AI Expected to ContinueAs the cost of computing comes down, private equity investment in artificial intelligence is expected to soar. Pull-through market technologies such as self-driving cars, digital assistants, and virtual reality gear will continue to drive new technology development. The Organization for Economic Co-operation and Development, (OECD) released a report in 2018 that studied global investment trends in AI technology. It states venture capitalists are “stepping up equity investments in artificial intelligence (AI) start-ups, reflecting a growing interest in AI technologies and their commercial applications.” CITATION OEC18 \l 1033 (OECD, 2018)After five years of steady increase, private equity investment in AI has accelerated since 2016, with the amount of private equity funding doubling from 2016 to 2017. In total, OECD estimates that “more than $50 billion was invested in AI start-ups during the period 2011 through to mid-2018. This surge in private investment suggests investors are increasingly aware of the potential of AI and are crafting their investment strategies accordingly.” CITATION OEC18 \l 1033 (OECD, 2018) FEMA Deputy Administrator Richard Serino notes corporate endeavors are “emerging in support of the development of new (AI) tools and opportunities for emergency managers. “For example, IBM has been working to combine its Watson technology with the weather data it acquired with its purchase of The Weather Company to enable third-party developers to experiment with a range of disaster-related solutions. Facebook has aggregated geolocation data from its users and provided it to assist humanitarian organizations after natural disasters. Google is using AI to develop enhanced flood warnings in India as part of its Google Public Alerts program. These solutions are agile and have the potential to reach millions of people with their light touch and mass-market appeal.” CITATION Ser18 \l 1033 (Serino, 2018)Traditional Newsrooms Find Hope In Artificial IntelligenceNewsrooms around the globe underwent a dramatic shift when news began to move online. Pocket-sized digital stories became the norm with longer in-depth pieces slowly dissolving into micro-blogs, podcasts, or other mobile-compatible formats. For some time, statistic-rich stories such as sports, weather, corporate earnings reports, and crime have been written by computers. Short, factual, and far from Pulitzer contenders, they get the job done. Robot reporters can be more thorough than human reporters as the Global Investigative Journalism Network found. Human journalists tend to single-source stories while AI-based software can import multiple sources of data, compare them, and identify trends and recognize patterns. CITATION Ron19 \l 1033 (Ronderos, 2019) Using Natural Language Processing (NLP), robot reporters humanize data into sophisticated sentences with adjectives, metaphors, and similes. Using data points from social media posts, they can even report on crowd emotions at an unusually close sporting event. CITATION Ron19 \l 1033 (Ronderos, 2019) While the rise of robot reporters may sound the alarm to many, a growing number of journalists seem to embrace the use of AI in the newsroom. According to the Global Investigative Journalism Network, AI could “become the savior of the trade— making it possible to better cover the increasingly complex, globalized and information-rich world we live in.” CITATION Ron19 \l 1033 (Ronderos, 2019)AI-based tools in the newsroom are adept at more than just churning through information and writing stories. They also are employed to fact-check. Reuters has employed News Tracer, a system that automates news production using Twitter data to identifying emerging conversations from more than 12 million tweets per day. It then contextualizes the story with a summary and topic. CITATION Liu17 \l 1033 (Liu, et al., 2017) News Tracer helps identify news stories by looking for patterns in emerging tweets. AI-based image recognition tools help newsrooms identify objects, locations, and even recognize faces. For example, The New York Times uses an AI program called Rekognition to identify members of congress in photos. Wibbitz is a software program that automatically creates scripts and can produce short videos in rough form. CITATION Ron19 \l 1033 (Ronderos, 2019).How It Works: Data Gets Sucked In, a News Story Gets Pushed Out Automated journalism relies on four major components: available data, algorithms, machine learning, and natural language processing. The algorithm component is the engine and coordinating filter for the incoming information. Machine learning, as best described by the SAS Institute, a global leader in analytical software, is a data analysis method that has the ability to automate analytical model building; it is a branch of artificial intelligence based on the premise computer systems can learn by using data to identify patterns and make decisions with limited intervention by humans. (SAS Institute, 2019) Natural language processing is a tool that helps computers “understand, interpret and manipulate human language.” It incorporates elements from several disciplines, including computer science and linguistics. In short, machine learning works to fill the gap between human language and computer understanding. (SAS Institute, 2019) Figure 1. Graphic representation of automated journalism.In a 2016 paper published by the Columbia School of Journalism, Andreas Graefe breaks down available technology solutions from simple code that extracts numbers from a database, which are then used to fill in data field in pre-written templated news stories, to more sophisticated approaches that analyze data that would make stories more compelling narratives. She adds that “the latter relied on big data analytics and natural language generation technology and emerged from the data-heavy sports reporting and both major providers of natural language generation technology in the United States, Automated Insights and Narrative Science, began by developing algorithms to write recaps of sporting events automatically.” CITATION And16 \l 1033 (Graefe, 2016)Figure 2. Illustration of how automated journalism works. CITATION And16 \l 1033 (Graefe, 2016)SOURCE: Review ConclusionArtificial intelligence is a fast-moving disrupter. Fueled by faster computing power and increased private sector investment, organizations without the capacity to handle the forthcoming torrent of information will struggle. AI-based systems will soon produce information at such a rate and volume, non-AI entities like the joint information center will require AI-based systems of their own to keep up, as well as control the narrative. Advancement in artificial intelligence has now risen to become a national directive, as well as a prominent component in the future of emergency management, but not its potential to support large scale JIC operations. While a growing amount of research on automated journalism can be found, there is little specific research on how these AI systems may be adopted for use in the joint information center, making this research somewhat unique. Because large newsrooms and joint information centers bear many similarities in organizations, as well as function, understanding how they have leveraged the power of artificial intelligence to increase productivity and support decision-making provides a foundation from which to begin exploration. METHODOLOGY: CASE STUDY AND SURVEYThis thesis explores the implementation of AI-based content creation technologies to improve joint information center accuracy and efficiency by examining three cases of automated journalism implemented in three major newsrooms: The Washington Post, Associated Press, and The Guardian research method includes a review of published literature, technology white papers, and analysis of resulting media coverage. While the literature review helps establish the viability of said technology within a fluid and high-volume content producing operation, a standard survey was conducted among working public information officers and communications professionals to measure awareness of and general interest in the application of similar AI-augmented technology in the joint information environment. Survey population sampleThis study sought to gather a diverse population of professional communications professionals as possible, and in such a manner, the sample represents the total population as closely as possible. An eight-question standard survey was distributed to a nationwide population comprised of public information officers, public affairs officers, and public relations professionals, and emergency managers or specialists that often step into the role of public communicators. The potential sample was vetted by distributing the survey through professional organizations that represent professional public communicators.Those organizations were:2019 EMI Master Public Information Officer Program Cohorts I, II, IIIPublic Relations Society of America Defense Information School Public Affairs Officers American Association of Airport Executives Digital Media Summit LinkedIn (Online)The survey instrument was designed to identify each respondent in their specific role to ensure respondent qualification. The option to determine one’s position as “other” was included to catch respondents that may not be qualified to participate. Those responses designated as “other,” if complete, were recorded and included in this study, but are kept separate from the primary body of data. While primarily designed to be a quantitative survey instrument, opportunities for respondents to provide qualitative feedback were included. These opportunities were marked as optional, and unlike the qualitative questions, were not set as required responses. Incomplete or abandoned surveys were not considered qualified were discarded. Those respondents that did provide substantial qualitative feedback were invited to participate in an optional narrative interview.Case Study: The Washington Post’s HeliografIn September of 2016, The Washington Post launched its own AI storytelling technology to create included 300 short stories and alerts on the Summer Olympics in Rio de Janeiro and 500 articles around the 2016 U.S. election, generating more than 500,000 article clicks. (Moses, 2017). In its first year on the job, the AI-system, known as Heliograf, part of The Post’s Arc Publishing Platform, autonomously produced and published 850 articles. It also covers D.C- area high school football games, and along with creating stories, Heliograf generates its own tweets. CITATION Mos17 \l 1033 (Moses, 2017).Figure 3. Screenshot of an AI-generated tweet. CITATION Mos17 \l 1033 (Moses, 2017). ?2017 Washington PostSOURCE: Screenshot by authorIn 2016, The Washington Post was able to cover every major race on Election Day—every House, Senate and gubernatorial race in the nation—with the help of Heliograf. Lukas I. Alpert writes about the feat in a Wall Street Journal article, published on Oct. 19, 2016. Alpert explains that The Post’s idea was to use artificial intelligence to bolster the work of some 60 reporters already assigned to election coverage. The reporters focused their attention on covering the high-profile races and those expected to be pivotal, while AI filled in the gaps. CITATION Alp16 \l 1033 (Alpert, 2016)Using templates and pre-written previews, Heliograf automatically updated stories as the results came in. Human journalists would then added analysis and color, along with using geo-targeting to position stories at the top of the page when viewed by a reader from a specified area, customizing content, making it more engaging for the audience. Heliograf, according to Alpert’s article, would also alert the newsroom if a candidate lagged behind or if results concluded or started trending in an unexpected direction. CITATION Alp16 \l 1033 (Alpert, 2016)Case Study: The Associated PressIn 2014, the Associated Press (AP) began its digital transformation using news writing algorithms, realizing it had to not only do more—it had to do it better. CITATION Sea17 \l 1033 (Stroh, 2017)The AP needed volume to enable more choice and satisfy additional needs from affiliate news organizations, CITATION Sea17 \l 1033 (Stroh, 2017) They also needed a way to differentiate themselves, and AI was the tool that enabled them to do both. CITATION Sea17 \l 1033 (Stroh, 2017)Stroh adds that before the AP’s partnership with Automated Insights, the company that powers their AI-enabled content management system, “an AP staff of 65 business reporters could write about 6 percent of earnings reports possible of America’s 5,300 publicly held companies.”Twenty-four months later, the AP’s AI system was able to write 3,700 quarterly earnings stories. CITATION Sea17 \l 1033 (Stroh, 2017) The AP estimates that artificial intelligence tools have freed up 20 percent of their reporters’ time spent covering stories on corporate earnings, and that AI is also moving the needle on accuracy. CITATION Mos17 \l 1033 (Moses, 2017) In the case of AI-augmented financial news coverage by AP, the error rate in the copy decreased, even as the output of stories increased more than tenfold, said Francesco Marconi, AP’s strategy manager and AI co-lead. CITATION Mos17 \l 1033 (Moses, 2017) Marconi continues that in addition to generating news coverage, AI has helped the well-known global wire service extract hidden insights from data and improve the editorial workflow process by automating such tasks as tagging photos, writing cations for videos and even deploying AI-power cameras able to photograph angles not available to human journalists. CITATION Sea17 \l 1033 (Stroh, 2017)Figure 4. Associated Press news story written by robot reporter Case Study: Guardian Australia's Reporter MateIn late January of 2019, Guardian Australia published its first robot-written news using an AI-powered open source stem called ReporterMate. The creation of Nick Evershed, Guardian Australia’s data interactives editor, is an open-source application that gathers data, analyzes it, and then spits out a news-style report in natural language, complete with publication-ready graphics. CITATION Nic19 \l 1033 (Evershed, 2019) Evershed admits that ReporterMate is not an original idea, besides being based on one of his previous projects called DisclosureBot, it differs in the fact it is not a proprietary commercial service. Evershed feels that if news media want to control how AI technology is used within the industry, then an open-source construct is the best way forward. Evershed has made the source code for ReporterMare available at often get a bad reputation, stemming from malicious attacks on websites and the non-human traffic they generate on social media. While the use of bots in the newsroom is nothing new, bots like those created by Evershed have disrupted the news industry and forced them to tap into the raw information readily available on social media. Guardian Australia's DiscloureBot (@AusDisclosure), for example, tweets anytime an Australian politician or political party amends their donations or gift and business interest list. In also may include a link to the document. Figure 5. Sample of Guardian tweet. ?2019 The GuardianSOURCE: by authorAs with other AI-automated journalism tools in other newsrooms, ReporterMate has achieved similar boots in coverage, as well as efficiency, generating publication-ready stories complete with explanatory graphics.Figure 6. Sample of Guardian news story published Jan. 31, 2019 and written by ReporterMate, an artificial intelligence storytelling platform.SOURCE: . Screenshot by author.RESULTSPrimary Survey ResultsThe researcher distributed the first of two survey instruments to the selected population in late June of 2019. Responses were suspended Aug. 12, 2019. Because the survey was distributed electronically with a sharable link, the total population number is not quantifiable. However, there 42 responses with 33 of those responses qualified for inclusion in the study. Late, incomplete, or abandoned surveys were discarded. The population accepted for this research is 33, of which 31 have worked in an active joint information center and the remaining two in an exercise-driven joint information center.The population, comprised of professional public information officers and other public communicators, of which 51 % identified as full-time public information of public affairs officers with a third working at the city or county level, and 18% at the state or federal level. The population break-out is as follows:Figure 7. Respondent self-assessment of knowledge of AI-based content creation systems.Figure 8. Respondent self-assessment of knowledge of AI-based content creation systems.To further gauge qualification, the survey asked respondents to provide a self-assessment of their knowledge of artificial intelligence-based content creation systems. The results indicated of the 33 respondents, 24% indicated they had either a “strong” or “extremely strong” understanding of the concept of artificial intelligence, with 30% identifying as having a “basic understanding: and another 30% admitting “no understanding” of artificial intelligence or its application to create content.Figure 9. Respondent self-assessment of knowledge of AI-based content creation systems.Moving forward, the survey then asked a series of questions about individual comfort levels with AI-based technology performing specific tasks within a joint information center environment. These tasks included the autonomous writing of reports and press releases, social listening, social media monitoring, social media creation and autonomous posting of said social media, automated translation between language, and the independent engagement of special needs populations. A qualifying option of “human-in-the-loop” was added to the survey choices to identify those interested and open to AI technology, but with the caveat of human oversight.Figure 10. Degree of comfort with reports and press releases written by artificial intelligence.Figure 11. Degree of comfort with artificial intelligence engaging in social listening.Figure 12. Degree of comfort with social media content written and scheduled by artificial intelligence.Figure 13. Degree of comfort with social media posted by artificial intelligence.Figure 14. Degree of comfort with artificial intelligence directly engaging stakeholders and directing them to vetted resourcesFigure 15. Degree of comfort with artificial intelligence providing emergency information such as boil water alerts, road closures, evacuation information.Figure 16. Degree of comfort with artificial intelligence translating between languagesFigure 17. Degree of comfort with artificial intelligence directly engaging special needs populations.Overall, results from the preliminary survey illustrate a relative degree of comfort among the respondents when it comes to working with AI-based technology. While respondents seemed somewhat open to using the technology, it was evident that most are far more comfortable with the technology itself than they are relinquishing full message and publishing control to it. The high number of responses indicating a preference for human-in-the-loop (HITL) technology shows promise for future application in the joint information center, but at a more incremental and graduated pace.Secondary Survey ResultsResults from a second survey, which gauged experience in a joint information center, as well as whether or not the respondent finds the concept of an AI-augmented content management tool beneficial, revealed an overwhelming 100 percent interest. Additionally, two respondents indicated interest in being on the development team if the concept gains momentum. Figure 18. Would you find an AI-augmented content creation and management tool be of benefit in the joint information center?Some respondents completed the optional “Thoughts and Suggestion” box and provided additional ideas and consideration. They include:Such a tool should be able to interface with WebEOC.Using a standard (but configurable) content management tool would be helpful as it would allow for required training and proficiency before an emergency.The tool would need to be scalable and flexible to meet the needs of different sized JICs and emergencies.The process for writing and reviewing and approving communications within the JIC is challenging. The tool should include content review and approval capabilities.This tool could provide a good structure for setting up and organizing a JIC, as well as managing—and extracting value from—the volumes of incoming information generated during an emergency.A content management tool or software package that allows those working on a JIC to communicate, interact, store, and generate information is a great idea since many times, communications within the JIC can be difficult. The large newsroom and the joint information center are close cousins. To survive, newsrooms have bet their future on AI and AI-based news writing “robots” to augment human capability, improve efficiency, decrease errors, and reach specialized populations. Joint information centers, also faced staffing challenges, increasing amounts of information to parse and the requirement to write and produce formulaic products echo a similar need for data in and written content out. Findings from this study include:Significant advances in AI content creation and management technologies are readily available.Private-sector use of AI-based news writing algorithms (robot reporters) has improved news coverage, reduced errors, and identified critical information trends that may have otherwise gone unnoticed.AI technology is already a significant component in the emergency management field, but there are no components that support the joint information system.Public information officers indicate they are comfortable working with AI and AI-based technologies, but are not ready to relinquish content creation or total control of public information messaging, yet. Human-in-the-loop protocol can help.Public information officers indicate they use a variety of tools to accomplish their mission. In addition to WebEOC, multiple tools such as Google, Trello, Slack, TweetDeck, and a variety of content management and web tools such as Drupal and Joomla are used—creating a fragmented and very siloed flow of information. The information gleaned from these tools still must be extracted and repackaged by hand, by the public information officer before release.Public information officers have expressed enthusiastic support for the development of an AI-based content creation and management tool—particularly one that would interface with WebEOC.Cost, politics, and agency liability rank among the top concerns with deploying and using AI technologies in the emergency public messaging environment.RECOMMENDATIONS This thesis offers three case studies demonstrating how major newsrooms have successfully integrated AI-based news writing algorithms—automated journalism tools—into their workflow. The findings of these case studies present some convincing evidence for the application of similar AI technology in the joint information center. While the researcher does not proport AI to be an all-inclusive solution to improving JIC performance, it does offer a pathway for further consideration, investigation, and application.RecommendationsRecommendation 1: Joint Information Centers should be looking at private-sector successes, such as major newsrooms, for ideas to organize and modernize operations. Automated journalism, fueled by social media, will exponentially increase the amount of information created before, during, and after an emergency. Existing joint information center operations will require more scalable and agile processes to remain relevant to audiences and ahead of the non-official narrative. Recommendation 2: Development of as standard, yet scalable content management system for the joint information center, based on automated journalism technologies. The system’s user interface would be designed on standard newsroom content management systems where documents are shared, and published either directly to a JIC human editor or be automatically published. Additionally, this package would include a user-configurable dashboard to quickly deploy and organize the joint information center, from scheduling to resource requests, documentation, and archival of work products created, to name a few. This AI platform, as identified nearly unanimously by the secondary survey results, should be directly able to access and data and events from WebEOC master events logs. The concept would be to create information unity and efficiency among multiple sources, much like a working newsroom.Recommendation 3: Develop public-private partnerships among FEMA, DHS Science and Technology Directorate and private-sector AI automated journalism companies, as well as the users and developers of WebEOC to form an exploratory product development team to discuss a way forward for a joint information content creation and management tool.DISCUSSIONAdvances in the application of artificial intelligence have made significant strides over the past decade, entering almost every facet of the global economy, as well as everyday life. As the next decade opens the doors to self-driving cars, self-flying planes, encrypted currencies, city planning, environmental stewardship, virtual surgery, and among many, writing news, the world sits on the edge of an unprecedented shift in the value of data. For status quo organizations that cling to the traditional “wait and see” attitude, they may find it much harder to catch up. With that in mind, the researcher feels it is of value for the public information field, its professionals as well as the agencies they support—and those that train them— to make an effort to understand how artificial intelligence will shape the joint information center of the future. Emergencies may not change, but how the news agencies generate information, and how stakeholders receive it will.Along with communication, high-volume information scrubbing will become essential, as well as the ability and extract relevant data from the information available will become a necessary skill. Large international newsrooms, with shrinking budgets, are now leveraging the power of artificial intelligence to power automated journalism. Tireless, these AI-based robot scribes have helped redefine newsroom workflow and productivity. Joint information centers, which share many traits with newsrooms, could learn from recent newsroom technology shifts; thus the researcher poses an exploration of how AI-based content creation—automated journalism—tools used in newsrooms may be adapted to improve general information messaging and joint information center operation.One of the most cumbersome challenges with this project was the fact, little scholarly research specific to the application of artificial intelligence exists. Sources within the literature review are derived from technology or emergency management sector publications, white papers, and from actual news coverage of the systems from the owning newsroom or other credentialed news entities—in some cases, the actual software creator. The researcher’s goal is to provide the reader with a basic understanding of automated journalism and the movement to make AI a national priority, then walk them through three case studies to illustrate successful implementation. The researcher then asks the reader to explore the concept of an AI-augmented tool for use in the joint information center environment. The topic’s broad perspective and the multitude of definitions made it challenging to communicate the concept and word survey questions. The researcher identifies the following weaknesses in this thesis:Significant lack of existing academic research or case studies on artificial intelligence used in the joint information center environment.The subject of artificial intelligence, itself, may have intimidated the prospective survey population resulting in many not attempting to complete the survey at all. Analytical software associated with the survey link indicated 248 pageviews with only a 13.7% conversion rate A larger population would have yielded far better results.The continuing advancements in technology presented challenges in researching, as well as finalizing recommendations. While the researcher does offer specific recommendations, they are provided in a conceptual form rather than supported with particular technologies, hardware, and software. While specific potentials and limitations of AI-based technologies are identified within this thesis, it lacks a rigorous discussion of each. To include such would require a subject less conceptual and one with more specified technologies, hardware, and software.Table 1Comparison of benefits and challenges of AI-augmented JIC, prepared by the authorPromising Benefits for AI-Augmented JIC Challenges, Limits for AI-Augmented JIC Data automationRoutinely scrubs multiple data sources for specific details, looks for patterns, anomalies, powers decision making.Expand fact-checkingThe technology available, but no particular applications for JICThe technology and tools are employed in the private sector, and some aspects of emergency management, but no movement toward proving tools for ESF 15 or JIC.Task automationUses templates to generate reports, fact sheets, social media analytics, briefings, updates, expenditures, calculates personnel costs, media analysis reports.Development costCosts to license existing technology. The Washington Post sells Heliograf’s technology through Arc Publishing, which starts at $10,000 per month. CITATION McC17 \l 1033 (McCoy, 2017) Open-source software programs are available but would require funding to develop.Helps shape a standard workflow for JIC operationPIOs arrive at a JIC, each with a toolbox. Duplicative efforts waste time and risk error. A standard (yet scalable content system) can improve operational performance.Program sustainability costOnce developed, the funding required to sustain the program, software integrity, initial set up, and compatibility with OEM partners.Image recognitionIdentifies locations, persons, structures, and events. Aids in the discovery of emerging problems or emergencies. TrainingA substantial training program would need to be considered.Social listeningAble to detect emerging conversations, concerns, inaccuracies. Identifies and can help pre-empt disruptive threads.The legality of data useCareful vetting of proprietary data should be considered. Although the AI program may be an open-source or public domain entity, running proprietary data (subscription-paid) through it may pose legal mentary moderationAble to detect emerging conversations, areas of concern, improvement, as well as mitigate inflammatory, inappropriate discussions and imagery, as well as to detect other bot activity. Identifies and can help pre-empt disruptive threads.Standard cyber issuesAs with any software system, the general site and network security concerns ensue.Content creationCan create new stories, press releases, reports, social media posts, limited videos, social media art. Can recommend Prepare for human review.Program sustainability costOnce developed, the funding required to sustain the program, software integrity, initial set up, and compatibility with OEM partners.Automates event documentationCan create, as well as automates document storage, back up and archival of JIC activities and work productsTraining and familiarizationA substantial training program would need to be considered. Research indicates there is a knowledge gap, as well as trust issues among PIOs and AI.Translation between languagesCan easily translate between multiple languages, to include regional dialectsThe legality/privacy of data useCareful vetting of proprietary data should be considered. Although the AI program may be an open-source or public domain entity, running proprietary data (subscription-paid) through it may pose legal concerns.AccessibilityAI may assist in creating inclusive designs—rather than expecting people to provide something in an accessible format; everything may become accessible in the future.Agency liabilityWho is at fault and who pays in the event something goes wrong? Multiple possibilities from the end-user to the software developer. Special needs populations and disabled personsVoice recognition, text-to-speech, and biometric login for those working in the JIC as well those seeking assistance from emergency resources.Reduce/identify gender, ethnic, religious biasAI algorithms can identify and help reduce gender, ethnic, and religious bias in public communication products and activities.Facilitates transparencyAutomated tasks can show a transparent chain of custody for information, documents, and information. Digital forensics helps identify information sources.Speeds up open records requestsSearchable documents and work products can be easily retrieved, screened, and perhaps even redacted for release to the media and public. Better tracking of open records requests.Suggestions for Future Work and ResearchRecommendations for future work on this topic would include:Development of a working, small-scale, open-source prototype tool to facilitate the evaluation of potential JIC applicationExploring available grant funding for joint information center technology and technology training programsRe-evaluation of current PIO skillset to build leaders for the data-driven JIC of the futureCONCLUSIONThe potential for customized AI-based content tools in the joint information center offers many promising benefits, as well as some new—and familiar challenges. Technology investments within the private sector and emergency management field are increasing, but investment in compatible tools for joint information center operation is minimal. Public information officers report to the joint information center with individual toolboxes of proprietary applications. In many cases, they use a tool such as Google Docs in an attempt to unify themselves and create a quasi-content management system. Private sector case studies have proven successful with the implementation of AI-based content creation and management tools. Survey results from this research indicate public information officers have a keen interest in a digital transformation that could, possibly, include similar AI-based tools. Whether this is a project best suited to a private entity or perhaps a public-private partnership would be best suited for this, is yet to be determined. However; if this thesis does nothing more than raise awareness to and open a discussion exploring the need for a specialized content management tool that can be deployed into joint information centers, much like WebEOC, and based on AI-tools used in newsrooms, the researcher will consider it a success.REFERENCES BIBLIOGRAPHY \l 1033 Alpert, L. I. (2016, October 19). The Wall Street Journal. Retrieved August 27, 2019, from Washington Post to cover every major race on election day with help of artificial intelligence: , J. (2018, May 24). These tech problems could hurt the government for years. Retrieved June 10, 2019, from Federal Times: , B. (2019, May 9). Artificial Intelligence. Retrieved July 22, 2019, from Encyclop?dia Britannica: , E. (2019, February 6). Meet Bertie, Heliograf and Cyborg, the new journalists on the block. Retrieved July 6, 2019, from Forbes: , E. (2015, August 13). Governments could pay high price for lagging on tech. Retrieved May 6, 2019, from Government Technology: Blankespoor, E. d. (2017, August 14). Capital market effects of media synthesis and dissemination. Review of Accounting Studies, Forthcoming, 1-2. Retrieved May 30, 2019, from , N. (2019, January 23). Why I created a robot to write news stories. Retrieved from The Guardian: , A. (2016). Guide to automated journalism. Columbia School of Journalism.Kazuo Yano, P. I. (2017). Artificial intelligence for happiness of people. Liu, X., Nourbakhsh, A., Li, Q., Shah, S., Martin, R., & Duprey, J. (2017). Reuters Tracer: Toward Automated News Production. Retrieved August 22, 2019, from , J. (2017, November 12). Competent robots are here. Retrieved August 29, 2019, from Content Marketing Institute: , L. (2017, September 14). The Washington Post's robot reporter has published 850 articles in the past year. DigiDay. Retrieved May 30, 2019, from . (2019, July 21). NSF joins federal partners in announcing update to national AI research and development strategy. Retrieved July 24, 2019, from National Science Foundation: . (2018). Private equity investment in artificial intelligence. OECD Going Digital Policy Note. Paris, France: Organisation for Economic Co-operation and Development. Retrieved April 10, 2019, from , W. (2014, March 17). The first news report on the L.A. earthquake was written by a robot. Slate Magazine. Washington, D.C., U.S. Retrieved Jun 6, 2019, from , M. T. (2019, January 22). How innovative newsrooms are using artificial intelligence. Global Investigative Journalism Network. Retrieved August 22, 2019, from Institute. (2019). Natural Language Processing (IoT). Retrieved August 24, 2019, from SAS: Institute. (2019). What is machine learning, and why does it matter. Retrieved August 21, 2019, from SAS: , R. (2018, September 29). The role of AI in preparedness: practical advice from former FEMA deputy administrator. Homeland Security Today. Retrieved August 3, 2019, from , S. (2017, December 1). As AI technology advances, so does its ability to assist journalists. Retrieved from Editor and Publisher: , L. (2014, July 1). Need to write 5 million stories a week? Robot reporters to the rescue. Mashable. Retrieved May 30, 2019, from ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download