Proposal Peer Review Rubric Criteria Abstract Is the abstract ~150 words or less? As the reader, summarize the purpose of the document & the change that you obtained from the abstracts. Is the justification for the change appropriate? Specify your view on the the change. Introduction/Background – Is the system and the proposed changes clearly defined? Specify any improvements would you make. – Is the background focused on only giving the information necessary? Specify any additional aspects you think should be included or removed. Audience/Tone Based on your reading, what is the tone of the proposal? What changes do you recommend to ensure a professional tone written to the appropriate audience? Body of Proposal – Is the problem with the current system clearly defined & justified? What improvements should be made. – Are the proposed changes clearly defined and well motivated by the problem statement? Specify improvements which can be made. – Do you believe it is plausible that the proposed changes could solve most, if not all, of the problem statement? i.e. was the problem statement concise and exact? Any suggestions? – Is it the market size for this system clearly outlined? Are statistics reported from credible market sources(Gartner, Forrester, IBIS World)? Provide Market Journal Database Link Comments Criteria Benefits/Cost/Schedule/etc – Are the benefits both desirable and convincing to the audience of this proposal? – Are the benefits of the proposed changes backed up by some amount of research, where necessary? – Are the costs and schedule plausible, given the writer’s background? (i.e. was any effort put into coming up with good figures?) Are there sufficient details specific to the project included in the schedule (ie. not a copy/paste Agile or waterfall schedule)? – Are the members of the team listed with appropriate qualifications? Does their professional background and level of skill meet the scope of the technical work outlined in the schedule? Overall Feedback Your overall critique of the review Comments PROPOSAL: AI-Driven Booking Ecosystem PREPARED FOR Matt Goldberg TripAdvisor PREPARED BY Fatih Sen November 5, 2023 1 AI-Driven Booking Ecosystem Proposal Abstract TripAdvisor is in the perfect position to fill in all the gaps in the current online travel market. For TripAdvisor evolving into a holistic AI-driven booking ecosystem that seamlessly integrates personalized trip plans and suggestions, is recognizing the platform’s present limits in responding to customer preferences and the absence of uniform booking capabilities. This solution suggests a paradigm change. TripAdvisor can improve user happiness, engagement, and loyalty by employing AI for personalized travel planning and implementing a one-stop booking platform. The theoretical underpinning, advantages, and practicality of integrating AI elements are highlighted in the technical part. According to research findings, AI-assessed data has the potential to improve income. The market size section makes clear the insights of the travel sector, which is expected to increase significantly. The suggested methodology details a staged implementation strategy, with the management portion outlining the expert team and budget, assuring a successful implementation. 2 AI-Driven Booking Ecosystem Proposal Table of Contents ABSTRACT…………………………………………………………………………………….. 2 INTRODUCTION……………………………………………………………………………… 4 System Overview………………………………………………………………………. 4 Problem……………………………………………………………………………………. 4 Solution……………………………………………………………………………………. 5 THEORETICAL SECTION………………………………………………………………… 5 Theory……………………………………………………………………………………….5 Benefits……………………………………………………………………………………..6 Feasibility…………………………………………………………………………………. 9 Market Size……………………………………………………………………………….. 9 Methodology…………………………………………………………………………… 10 MANAGEMENT SECTION………………………………………………………………. 11 Team and Budget…………………………………………………………………….. 11 Schedule…………………………………………………………………………………. 11 CONCLUSION………………………………………………………………………………..12 Works Cited………………………………………………………………………………….. 13 3 AI-Driven Booking Ecosystem Proposal Introduction System Overview TripAdvisor is a well-known and trusted online and mobile travel platform that provides travelers with information, reviews, and recommendations on all aspects of travel. From hotels, restaurants, attractions, and more. To effectively plan their trips, users can browse a large database of both official and user-generated content, browse photos, and read detailed reviews. Aside from its review-centric features, this user-driven platform has become a go-to destination for travelers worldwide, providing valuable insights and guidance for a wide range of travel experiences, making it an indispensable tool for both leisure and business travelers. Problem TripAdvisor though comprehensive in its methodology and approach towards displaying a breadth of activities and locations across the world for its users to choose from, doesn’t approach users by catering to them with personalized itineraries and travel packages based on their experiences. On top of this, it also lacks the ability for its users to take the next steps into unifying and booking all their travel decisions on the platform itself. 4 AI-Driven Booking Ecosystem Proposal Solution TripAdvisor’s evolution into a comprehensive one-stop platform for booking flights, accommodations, activities, and transportation, seamlessly integrated with AI-driven personalized travel itineraries and recommendations tailored to each user’s preferences and past booking experiences, can cover all of the gaps that can currently be closed on the platform in terms of increasing user engagement, loyalty, and revenue potential. Technical Section Theory A unified booking system based on AI-driven travel recommendations and plans pivots the focus of TripAdvisor away from being just a reference point for the user in their travels. This makes way for personalization, making TripAdvisor more versatile. A study on the relationship between AI and user satisfaction emphasizes that “tourists’ trust in the big data and AI-based PTR system increased with perceived personalization, visual appearance, and information quality.” (Yang, Xinran, et al). The current state of the application ignores user needs, cutting off important details that 5 AI-Driven Booking Ecosystem Proposal users want to be exposed to. Implementation of AI recommendation and booking is a critical missing feature of the app. Benefits Augmenting AI-driven personalized booking options will enhance user experience, merge and ease the booking process, and increase customer loyalty and result in three main proven benefits: 1. Making TripAdvisor cater to all the user’s booking needs for their personalized plans will increase user satisfaction making it a preferred platform for users. We can enhance user engagement through a better user experience just as been clarified in a rigorous study, “Over three-quarters of consumers (76 percent) said that receiving personalized communications was a key factor in prompting their consideration of a brand, and 78 percent said such content made them more likely to repurchase.” (Arora et al.) Imagine a traveler receiving an intelligently curated itinerary based on their preferences, interests, and past experiences. This level of personalization won’t only boost satisfaction but also foster loyalty to TripAdvisor. In another study it was found that “tourists’ trust in the big data and AI-based PTR system increased with perceived personalization, visual appearance, and information 6 AI-Driven Booking Ecosystem Proposal quality.”(Yang et al. ) Travelers need assurance that the recommendations provided are both accurate and trustworthy, and this trust can be built through personalization. Building on these insights, it’s essential for TripAdvisor to refine its personalization strategies continually. The platform can further tailor recommendations and content to match each traveler’s unique preferences by harnessing the power of AI and its fine-tuned models. 2. Making TripAdvisor the preferred booking platform will inevitably increase loyalty and preference for TripAdvisor’s system. Based on insight from a recent study it was discovered that “when users believe that using MHB (Mobile Hotel Booking) technology fits well their lifestyle and fits well the way they like to book a hotel room, then they find MHB convenient and easy to use and thus are willing to continuously use MHB and recommend it to others.” (Ozturk et al.) By integrating flights, accommodations, activities, and transportation into a one-stop platform, TripAdvisor can address the inconvenience that users often find elsewhere. This convenience will not only attract new users but also keep them coming back. Researchers have stated: “Our results provided empirical support for the existence of all main effects (H1, H2, H3, H4, and H5 were supported), thereby 7 AI-Driven Booking Ecosystem Proposal confirming that monetary, quality-of-benefits, social status, information, and preference values share a significant positive association with purchase intention toward OTAs.” (Talwar et al.) TripAdvisor’s comprehensive approach can boost conversion rates, translating to higher revenues. A one-stop platform for all travel needs will improve the user experience and incentivize more bookings. The outcome will ultimately be growth and a cycle of recurring users building an ecosystem of happy travelers who left satisfied and looking forward to their next experience booking a trip with TripAdvisor. 3. Revenue will see an increase through the data that AI will be assessing. Leveraging user-provided data and history for informed personalized marketing will increase revenue as reported “78% of organizations that follow a data-driven approach verify an increase in lead conversion and customer acquisition.” (Awan) By utilizing user data, TripAdvisor can deliver highly targeted and effective marketing campaigns. This not only benefits the business but also ensures users receive promotions and offers that are genuinely relevant to their preferences. In another study “It was found that hedonic and utilitarian features are important variables influencing future consumer decisions through relationship marketing.”(Bilgihan, et al.) And so we have the opportunity to create an emotional connection with travelers by offering 8 AI-Driven Booking Ecosystem Proposal experiences tailored to their desires, making their journey more enjoyable and memorable. Feasibility With the guidance of our Director of Engineering and our proficient team of senior engineers, all well-versed in the standards and requirements of the TripAdvisor software, we can seamlessly enhance the application’s features in an intuitive manner without causing any disruption to the familiar user experience. Users will continue to operate the app as usual, but now with the added capability of booking their desired travel destinations directly through the application. Our Director will oversee the coordination of communication with hotel systems and bookings, ensuring a successful integration through methodologies similar to those employed by online travel agencies. Leveraging the expertise of our senior engineers, particularly in the field of generative AI and its integration, we are confident that this transition will be executed promptly and efficiently. Market Size The travel industry is undergoing a transformative shift, particularly with the rise of the sharing-based economy and the increasing prevalence of smartphones. According to IBIS World, this trend is evident in the ridesharing transportation market. 9 AI-Driven Booking Ecosystem Proposal In a parallel development, the online travel industry is projected to experience remarkable growth, with an estimated revenue of nearly $1.5 trillion anticipated in the next four years, as reported by industry experts. Methodology Our suggested strategy combines collaborative efforts headed by our Director of Engineering and a team of distinguished engineers to integrate AI-driven recommendations and booking systems into TripAdvisor smoothly with a partnership with OpenAI. We place an emphasis on thorough market research and user analysis to inform the adoption of AI capabilities, beginning with an A/B testing program and gradually introducing the booking system. Our strategy relies heavily on effective connection with hotel systems, user-friendly interface design, and rapid development and deployment by our professional engineers. Continuous monitoring and improvement, made possible by user feedback loops, ensures that the evolving system remains responsive to user needs and market trends. This strategic approach ensures the integration’s practicality and promotes TripAdvisor as an innovative leader in the online travel market. 10 AI-Driven Booking Ecosystem Proposal Management Section Team and Budget The salaries arrived at in the table below are based on Glassdoor salary estimates for like positions in Silicon Valley. The total is from 24 40-hour weeks. Position Organization Name Hourly Salary Total Salary Director, Software Engineering Hopper Michael Curtis $300 $288,000 Principal Software Engineer AirBnB Sean Glover $250 $240,000 Senior Software Engineer TripAdvisor Joey Engelhart $170 $163,200 Senior Software Engineer TripAdvisor Yash Gazula $170 $163,200 3rd Party AI Models OpenAI Total: $146,000 $1,000,000 Schedule Implementation will be broken down into four phases. Weeks Phase Description 1-4 Thorough examination and documentation of system needs All members alongside the collection of necessary AI and booking resources, to lay the foundation for subsequent development. Requirements Analysis and Resource Collection Team Members 11 AI-Driven Booking Ecosystem Proposal 5-11 Design and Prototyping System’s architecture is meticulously designed, and prototype models are created to visualize and validate the proposed structure before full-scale implementation. 12-20 Implementatio Software development takes place to bring the envisioned n features to life. 21-24 Testing and Deployment Engineers All members Rigorous testing to identify and rectify any issues, following Engineers which it is deployed for operational use. Conclusion The transformation of TripAdvisor into a comprehensive one-stop platform, integrated with AI-driven personalized travel itineraries and recommendations, holds immense promise. By addressing the current limitations in the industry, we can enhance user engagement, satisfaction, and revenue potential. It is an opportunity to redefine the travel industry and secure TripAdvisor’s position as a market leader. By harnessing AI-driven personalized travel experiences, TripAdvisor can not only redefine the travel industry but also significantly increase user engagement, satisfaction, and revenue, ultimately securing its position as a market leader. Our budget and team members make it highly feasible to embark on and complete this transformation within the 24 weeks appointed. 12 AI-Driven Booking Ecosystem Proposal Works Cited Arora, Nidhi, et al. “The Value of Getting Personalization Right-or Wrong-Is Multiplying.” McKinsey & Company, 12 Nov. 2021, www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/the-v alue-of-getting-personalization-right-or-wrong-is-multiplying. Bilgihan, Anil, and Milos Bujisic. “The Effect of Website Features in Online Relationship Marketing: A Case of Online Hotel Booking.” Electronic Commerce Research and Applications, vol. 14, no. 4, 2015, pp. 222–232, doi:10.1016/j.elerap.2014.09.001. Global Business Travel Industry Forecast Is for Accelerated Rebound, Spending to Reach $1.8 Trillion by 2027 – Global Business Travel Association.” GBTA, Global Business Travel Association, 23 Aug. 2023, www.gbta.org/global-business-travel-industry-forecast-is-for-accelerated-rebo und-spending-to-reach-1-8-trillion-by-2027/. Muhammad Bilal Awan. “Data-Driven Marketing for Better Roi.” Data Science Dojo, 27 Oct. 2022, datasciencedojo.com/blog/data-driven-marketing/. Ozturk, Ahmet Bulent, et al. “What Keeps the Mobile Hotel Booking Users Loyal? Investigating the Roles of Self-Efficacy, Compatibility, Perceived Ease of Use, 13 AI-Driven Booking Ecosystem Proposal and Perceived Convenience.” International Journal of Information Management, vol. 36, no. 6, 2016, pp. 1350–1359, doi:10.1016/j.ijinfomgt.2016.04.005. Talwar, Shalini, et al. “Why Do People Purchase from Online Travel Agencies (Otas)? A Consumption Values Perspective.” International Journal of Hospitality Management, vol. 88, 2020, p. 102534, doi:10.1016/j.ijhm.2020.102534. Yang, Xinran, et al. “Personalized Tourism Recommendations and the E-Tourism User Experience.” Journal of Travel Research, 2023, doi:10.1177/00472875231187332. 14 Proposal for: Customized “For You Page” Prepared by: Kevin Yu System Name: TikTok Abstract: In modern times, the internet has increasingly gained popularity, and social media has taken over the world. Its impact increases every year, and it influences how people think and learn. TikTok, the first social media platform involving the scrolling of short one to three minute videos, revolutionized how social media content is viewed and consumed today. The platform uses an algorithm to determine what videos to show users on the “For You Page”, and every user of TikTok adheres to this algorithm. Although the algorithm is effective for many users, it is not always effective for others as it constantly shows videos with different topics and lengths, which users may not like. To combat this, Arman Khondker, Kaixin Xiao, and I are proposing that TikTok should create a customizable “For You Page” where users select specific topics in order to be fed a consistent stream of content. Users may have access to both the original “For You Page” to see new trending content, and they will also be able to view their own “For You Page” which would feature content they select. 1 Table of Contents Abstract.………………………………………………………………………………………….1 List of Figures.……………………………………………………………………………………2 Introduction.………………………………………………………………………………………3 System Overview………………………….…………..…………………………………3 Problem.…………………………………………………………………………………..3 Solution.…………………………………………………………………………………..3 Technical Section ……………………………………………………………………………..…4 Theory.……………………………………………………………………………..…….4 Benefit.……………………………………………………………………………………4 Feasibility.……………………………………………………………………………..…6 Market Size.………………………………………………………………………………6 Method of Approach……………………………………………………………………..6 Management Section.…………………………………………………………………………….7 Team………………………………………………………………………………………7 Schedule………………………………………………………………………………….7 Budget…………………………………………………………………………………….8 Conclusion……………………………………………………………………………………….8 References……………………………………………………………………………………….10 List of Figures Figure 1: Table of Team Members……………………………………………………………….7 Figure 2: Table of Project Schedule………………………………………………………………7 Figure 3: Table of Salary Costs…………………………………………………………………..8 2 Introduction System Overview: TikTok is a social media platform that allows users to see other users’ short videos and allows users to upload their own short videos. Users will be subject to short ten seconds to three minute videos that the user can constantly scroll through. The videos are shown on a page called the “For You Page”, and they are determined by an algorithm that decides what the user will see next. Overall, Tiktok has over one billion worldwide users from around the world, and according to Statista, it accounts for 20% of internet users worldwide. Problem: In TikTok’s current “For You Page”, videos are shown to users by the use of an algorithm. As a result, users get constantly changing videos of varying topics that they may or may not like. For example, users may be shown the latest trend going around the world. Although TikTok has a “Following” tab which shows users videos of content creators that they follow, it does not show any new content from different creators in the same topic or field. The “For You Page” and “Following” tab both have their issues. One is too specific while the other is too broad. Solution: By adding a customizable and separate “For You Page” that does not run on the normal TikTok algorithm, users of TikTok may input their own content preferences, leading to a more consistent stream of content that they will like. Users will be able to explore and discover new content creators and content of similar topics. The result is a happy medium that features new content in a specific topic range. As a result, the new feature would not have a super wide range of content, and it is also not limited to specific content creators. 3 Technical Section Theory: By adding a customizable and separate “For You Page” that does not run on the TikTok algorithm, users of TikTok may input their own content preferences, leading to a more consistent stream of content that they will like. As a result of the personal feature, TikTok will be able to send specialized ads generating more income, increase its usability raising user loyalty, and increase customizability leading to more users. Benefits: Adding this customization feature would not only help TikTok earn more revenue, but it would also allow TikTok to keep and attract more users to the platform. 1. Adding a customizable “For You Page” TikTok may target specialized advertisements towards their user, generating more income for the company. A study conducted by Jiang and Wu compared the benefits of targeted advertising and mass advertising. In its conclusion the study stated, “This result indicates that the targeted advertising might induce intensified competition. In all, we show that the most efficient free-entry outcome occurs when the targeting precision is moderate. In other words, neither mass advertising nor perfect advertising leads to a less efficient outcome” (Jiang and Wu). The reason that moderate targeting precision was deemed the most effective was due to the money in to money out ratio. As companies spent more money finding user’s interest, the money produced back from the advertisements eventually staggered and dropped, which is the reason that a moderate targeting precision level was deemed optimal. In the case of TikTok, the company would be able to save money on advertising 4 precision because users are already selecting topics they like. As a result, they would have low targeting costs as well as higher profits. 2. A customizable “For You Page” adds usability for the application and will lead to higher user satisfaction. A study conducted by Dianat et al. showed this direct correlation, it stated, “User satisfaction was also influenced by Web design attributes… Among the web design attributes, Web structure showed the strongest association with user satisfaction, followed by layout, personalisation, search and performance” (Dianat et al.). Adding the customizable “For You Page” to TikTok adds to the platform’s usability, which would increase the overall user satisfaction of the platform. 3. Adding customizability and differentiation leads to a more successful platform by raising user loyalty and attracting more users. As stated in a study analyzing application innovation and success, “Customers will look for products from other companies they feel like can satisfy them. For this reason, continuous innovation is needed”(Farida and Setiawan). Implementing the customizable “For You Page” aids in the innovation of the platform, keeping the user satisfied and engaged. In addition to a more satisfied user, “If product or service is unique, this strategy provides high customer loyalty” (Mustafa and Islami). By adding the customizability feature, TikTok will differentiate itself from its competitors, and it will keep their users while also appealing to other social media users looking for specific features. 5 Feasibility: Adding a separate tab for a customizable “For You Page” is very feasible as they have added the “Following” tab and the regular “For You Page” tab. The customizable page is very similar to the other two, and it will only feature a different algorithm. With the knowledge of TikTok’s platform and the backend development expertise of Arman Khondker, the implementation of the customizable “For You Page” will be a swift and simple task. The only new screen that must be added is a tab where users can enter topics that they would like to see, which is an easy addition. The implementation of this screen will be handled by Kaixin Xiao, a frontend engineer at TikTok with vast experience in implementing tabs. A fully functional customizable “For You Page” would give the user the topic screen, and it would then operate like TikTok’s current “For You page. The videos that are chosen to be shown to the user will be displayed by a different algorithm. Market Size: According to Statista, the social media market is worth around 140 billion dollars, and it is growing each year. From 2022 to 2023 alone, the market increased by 12 percent, and it is projected to keep growing in the upcoming years. Out of the total market, DemandSage found that Tiktok makes up around 22.4% of all social media users on the internet. Adding the customized “For You Page” will add room for more growth and greater profits. Method of Approach: Arman Khondker will utilize his backend expertise and experience working with TikTok’s algorithm to lead the addition of a new algorithm of TikTok by editing and tweaking 6 parts of the previous algorithm. He will create the new algorithm, and he will then implement that algorithm with the existing user interface of TikTok. With Kaixin Xiao’s experience in front end development with TikTok, she will be in charge of reading user input for topics they would like to see. Kaixin will read and store users’ input and will pass the data to Arman, which he will use in his new algorithm. Management Section Team Name Organization Qualifications Arman Khondker TikTok Back-end engineer at TikTok. Worked directly with the TikTok algorithm. Kaixin Xiao TikTok Front-end engineer at TikTok. Experience in creating new tabs and screens. Fig 1: Table of Team Members Arman Khondker and Kaixin Xiao are ideal members for the job. They are extremely qualified and have the experience that perfectly suits the project. Both members have directly worked with the TikTok application, and they have had years of experience working in the field. Schedule Implementing the customized “For You Page” will take around 6 weeks. The project will follow an agile methodology approach and will feature 6 phases. Week Phase Description Team Members 1 Planning Discuss the method of approach. Add ideas for effective implementation. All team members 7 2 Designing Create and Implement the tab that reads and Kaixin Xiao and her outputs user input. team 3 Developing Tweak the current TikTok algorithm to suggest videos within a certain topic range that the user selects. Arman Khondker and his team 4 Testing Combine both part 2 and 3 and test if the full feature works properly. All team members 5 Deploying Add the feature to the app and record user feedback. All team members 6 Reviewing and Tweaking Make changes based on user feedback . All team members Fig 2: Table of Project Schedule Budget: Team members will work full time, 40 hour work weeks. According to Glassdoor, the total estimate of the project will be around $33,120. Position Name Hourly Salary Total Salary (40 Hours / Week) Backend Engineer Arman Khondker $85 $20,400 Frontend Engineer Kaixin Xiao $53 $12,720 Total $33,120 Fig 3: Table of Salary Costs Conclusion TikTok has a massive audience and influence in the social media industry. It has revolutionized the consumption of short form content, and it has changed the way people learn today. Adding a customized “For You Page” will add application usability, raise user satisfaction, 8 and separate TikTok from competitors. As a result of this feature, TikTok will attract more users while also generating more profits. 9 References Dianat, Iman et al. “User-centred web design, usability and user satisfaction: The case of online banking websites in Iran.” Applied ergonomics vol. 81 (2019): 102892. https://doi.org/10.1016/j.apergo.2019.102892 Ganesh Iyer, David Soberman, J. Miguel Villas-Boas, (2005) The Targeting of Advertising. Marketing Science 24(3):461-476 https://doi.org/10.1287/mksc.1050.0117 Ida Farida, Doddy Setiawan, Business Strategies and Competitive Advantage: The Role of Performance and Innovation, Islami, X., Mustafa, N. & Topuzovska Latkovikj, M. Linking Porter’s generic strategies to firm performance. Futur Bus J 6, 3 (2020). https://doi.org/10.1186/s43093-020-0009-1 Journal of Open Innovation: Technology, Market, and Complexity, Volume 8, Issue 3, 2022, 163, ISSN 2199-8531, https://doi.org/10.3390/joitmc8030163. Jiang, Z., & Wu, D. (2022). Targeting Precision in Imperfect Targeted Advertising: Implications for the Regulation of Market Structure and Efficiency. SAGE Open, 12(1). https://doi.org/10.1177/21582440221082132 Tsai, HT., Chien, JL. & Tsai, MT. The influences of system usability and user satisfaction on continued Internet banking services usage intention: empirical evidence from Taiwan. Electron Commer Res 14, 137-169 (2004) https://doi.org/10.1007/s10660-014-9136-5 10 TikTok Harm Detection Proposal Proposal for: Misinformation and harm Detection System Prepared by: Matthew Hoffman System name: TikTok Abstract TikTok’s short-form videos have led to it having a meteoric rise in popularity, taking a part of the daily lives of many. However, with this popularity, there have been well-known negative externalities. Misinformation, challenges, and even positive content about harmful products lead to both harm for users and negative attention from the public and government. Google co-founder Larry Page, YouTube co-founder Jawed Karim, and I are proposing a machine learning system which can detect this harmful content and alert the moderation team, which should reduce the spread of misinformation, prevent harm to users, and reduce outside calls for regulation. 2 Table of Contents ABSTRACT……………………………………………………………………………………….2 List of Figures……………………………………………………………………………………..3 INTRODUCTION…………………………………………………………………………………4 System Overview………………………………………………………………………….4 Problem……………………………………………………………………………………4 Solution……………………………………………………………………………………4 TECHNICAL SECTION………………………………………………………………………….5 Theory……………………………………………………………………………………..5 Benefits……………………………………………………………………………………5 Feasibility………………………………………………………………………………….6 Market Size………………………………………………………………………………..7 Method of Approach………………………………………………………………………7 MANAGEMENT SECTION………………………………………………………………………8 Team……………………………………………………………………………………….8 Schedule……………………………………………………………………………………8 Budget……………………………………………………………………………………..9 CONCLUSION……………………………………………………………………………………9 REFERENCES…………………………………………………………………………………..10 List of Figures Figure 1: Mockup of user view……………………………………………………………………7 Figure 2: Table of team members…………………………………………………………………8 Figure 3: Table of project schedule………………………………………………………………..8 Figure 4: Table of Budgets…………………………………………………………………………9 3 Introduction System Overview TikTok is a mobile and web-based video hosting platform specialized in short-form videos. To facilitate this, it provides a system for uploading videos on both the web and mobile versions, along with a feature for recording and making some edits to videos on the mobile version. Once uploaded, TikTok’s algorithm may show that video to users when they’re scrolling on the “For You” page, along with having the video available through a direct link and the creator’s page. Each video has a length limit of 10 minutes. Problem TikTok’s video style of having quick, flashy videos which reward attention-seekers contributed to its success. However, it is also conducive for the spread of misinformation, harmful challenges, and inflammatory content. These can be much shorter than their counterparts due to the content being more attention-grabbing and needing less time due to a lack of nuance. The resulting harm has both killed people and drawn public ire towards TikTok. Solution Implementing an automated machine-learning system to detect when a video contains undesired and harmful content and alerts moderation staff can help solve the issue. The system can be taught to detect misinformation, harmful challenges, or promotion of harmful behaviors. Whenever a video is uploaded, this system would scan the video, and on detecting any of the above, would temporarily freeze the upload until a moderator takes action. By preventing these from even showing up, the harm and bad attention can be greatly reduced. 4 Technical Section Theory Implementing an automated system to detect harmful content will easily allow for detection and removal of undesired content from the platform. Without it, TikTok presently allows for misinformation, harm, and outside ire. As such, its implementation is critical for the safety of the platform and its users. Benefits Implementing a system to scan videos for misinformation or promotion of dangerous actions and halt their upload can provide TikTok and its users with three overall benefits: 1. Preventing the upload of misinformation can prevent its spread, through both TikTok and overall society. Preventing misinformative videos from being uploaded means misinformation is spread less. A study on how social media impacts vaccine misinformation found that “the combination of social media and the wide availability of highly legitimized forms of misinformation has accelerated its diffusion” (Di Domenico, et al.). Social media is a very strong vector for the spread of misinformation, and TikTok, being potentially the fifth most widely used social media platform, is a major potential player in this field. As such, TikTok directly addressing misinformation can have a strong positive impact, cutting off a major route through which misinformation could otherwise spread. 2. Preventing the upload of videos promoting harm can increase user’s safety On top of the safety that comes from preventing misinformation that would lead to people avoiding medical treatment and vaccines, there’s also safety from preventing things that lead to 5 harm. A case study about the Benadryl Challenge, which started circulating on TikTok in 2020 and which has continued since, reported that “teens have presented to emergency departments in Canada and the United States in various states of intoxication after taking high doses of diphenhydramine because of these social media “challenges,”…[with] at least one reported death,” (Elkhazeen et al). This case study was even in 2022, before the 2023 case of a 13 year old dying from a seizure due to attempting this “challenge.” TikTok’s 40 thousand safety professionals were unable to prevent this challenge from continuing to spread even after several years, resulting in more life being lost, and this is far from the only challenge. Were the proposed system in place, that moderation staff would have been significantly more likely to stop the video which led to the user attempting this challenge and subsequently dying. 3. Preventing misinformation and harm will help TikTok prevent regulatory oversight. TikTok is familiar with governments attempting to regulate and ban its usage by this point. Part of what leads to governments regulating TikTok is public pressure, caused by the harm TikTok plays a role in spreading. A study about materials promoting e-cigarette usage on TikTok openly stated that “Government regulation that captures e-cigarette advertising, promotion and sponsorship, including on social media, is needed” (Jancey et al.). Material which harms users draws negative attention, and that negative attention is exactly why governments may seek to place regulations on TikTok. Some degree of effective self-regulation, which this autonomous system provides, is required for convincing both the public and the lawmakers that governmental regulation is not required. Feasibility Implementation of this automated system will be moderately difficult, but with the skill and knowledge of Google co-founder Larry Page and YouTube co-founder Jawed Karim, the 6 task will be approachable. The user won’t need any new pages on their end, with the only modification being an alert to when a video’s upload is on hold, and whether the hold is lifted or the video is outright prevented from uploading. The largest challenge is creating the system for machine learning, training it, and integrating with TikTok and its report and flagging systems. Market Size TikTok has a strong market hold, with Business of Apps reporting that TikTok has around 1.5 billion monthly users in 2023, and that its 2022 revenue is estimated to be $9.4 billion. Even with this wide user base, TikTok falls behind several other Figure 1: Mockup of user view social media platforms in terms of monthly active users. However, its user base is expected to grow to 2 billion by the end of 2024. Method of Approach Jawed Karim will apply his experience in designing aspects such as PayPal’s anti-fraud system to the design and planning of the system’s backend, while Larry Page will apply his experience creating and running Google to design how the system will interface with both users and moderation staff. Figure 1 depicts a mockup of what a user may see when uploading a video if the system detects harm or misinformation. Options have been grayed out to show that the upload has begun, but has been frozen. 7 Management Section Team Name Organization Qualifications Jawed Karim YouTube(former) Designed PayPal’s anti-fraud system, co-founded YouTube site, MS in computer science from Stanford University Larry Page Alphabet Former co-founder and CEO of Google and Alphabet, MS in computer science from Stanford University Figure 2: Table of team members Jawed Karim and Larry Page are both educationally qualified and heavily experienced in fields which are valuable for projects working together with a system such as TikTok. Schedule Implementation of the machine learning misinformation & harm detection system will take an estimated 20 weeks. The project will follow an agile development model, with most of those weeks set in cyclical 4-week periods. Weeks Phase Description Team Members 1 Initial planning Given objective, plan ideas for how it should work and what it must look out for All (1 team) 2 Initial modeling Model the processes, front and back end All (1 team) 3-6 Initial coding Implementation of code, following plan and model All (2 teams) 7, 11, 15, 19 Testing Test with QA and moderation teams Larry Page 8, 12, 16 Plan and model Plan and model changes based on tests All (1 team) 9-10, 13-14, 17-18 Further coding Implement changes to code in line with revised plan and model All (2 teams) 20 Deployment Deploy and integrate system Jawed Karim Figure 3: Table of project schedule 8 Budget Team members will work 40 hours per week. Figures are estimated using the high end estimate on Glassdoor, based on the TikTok LA office location, of $185k annual. Pay divided by the hours expected to be worked by someone working a half of a year, rounded up, for the hourly salary. Position Name Hourly Salary Total Salary Software Engineer Larry Page $180 $136,000 Software Engineer Jawed Karim $180 $115,200 Total $251,200 Figure 4: Table of Budgets Conclusion TikTok is a very powerful competitor in the social media market, but its current lack of strong automated moderation puts it at risk of being held back by its harm to users causing bad public relations and potential government regulation. The long-term benefits of implementing such an automated system will promote the product’s growth. 9 References Di Domenico, Giandomenico, et al. “Marketplaces of Misinformation: A Study of How Vaccine Misinformation Is Legitimized on Social Media.” Journal of Public Policy & Marketing, vol. 41, no. 4, 2022, pp. 319–35, https://doi.org/10.1177/07439156221103860. Elkhazeen, Abu, et al. “A TikTokTM ‘Benadryl Challenge’ death—A Case Report and Review of the Literature.” Journal of Forensic Sciences, vol. 68, no. 1, 2023, pp. 339–42, https://doi.org/10.1111/1556-4029.15149. Iqbal, Mansoor. “Tiktok Revenue and Usage Statistics (2023).” Business of Apps, Business of Apps, 31 Oct. 2023, www.businessofapps.com/data/tik-tok-statistics/. Jancey, Jonine, et al. “Promotion of E-Cigarettes on TikTok and Regulatory Considerations.” International Journal of Environmental Research and Public Health, vol. 20, no. 10, 2023, p. 5761–, https://doi.org/10.3390/ijerph20105761. 10
Purchase answer to see full attachment