サイバー空間におけるフェイクニュースの広がりとその対策

ICT-ISAC 大阪WG-Dayで講演した「サイバー空間におけるフェイクニュースの広がりとその対策」の資料です。

Transcript

1. αΠόʔۭؒʹ͓͚ΔϑΣΠΫχϡʔε ͷ޿͕Γͱͦͷରࡦ ଜࢁଠҰ ԣ඿ࠃཱେֶ ॿڭ .BJM
   NVSBZBNBUBJDIJCT!ZOVBDKQ ICT-ISAC ⼤阪WG-Day
2. 2 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥʢΞςϯγϣϯɾΤίϊϛʔʣ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
4. 5 ϑΣΠΫχϡʔεͷ൙ཞ എܠ બڍલϲ݄Ͱ໿ສճɺީิऀਓ τϥϯϓࢯ
   ΫϦϯτϯࢯ ͷϑΣΠΫχϡʔε͕'BDFCPPLͰ֦ࢄ Allcott, Hunt, and Matthew Gentzkow. "Social media and fake news in the 2016 election." Journal of economic perspectives 31.2 (2017): 211-36.
5. 6 ϑΣΠΫχϡʔεͷ൙ཞ എܠ 5G network spreads COVID-19 Vitamin C cures COVID-19
6. 7 ϑΣΠΫχϡʔε࡞੒ͷϝϦοτ എܠ https://www.newsweek.com/2017/06/16/big-data-mines-personal-info-manipulate-voters-623131.html ੈ࿦ૢ࡞ ͓ۚṶ͚ https://www.nbcnews.com/news/world/fake-news-how-partying-macedonian-teen- earns-thousands-publishing-lies-n692451
7. 8 ϑΣΠΫχϡʔεͷඃ֐ എܠ ๫ಈࣄ݅ ͓ۚṶ͚ https://www.npr.org/2018/07/18/629731693/fake-news-turns-deadly-in-india https://www.washingtonpost.com/news/worldviews/wp/2013/04/23/syrian-hackers-claim-ap-hack-that- tipped-stock-market-by-136-billion-is-it-terrorism/ גՁٸམ
8. 9 ϑΣΠΫχϡʔεͷඃ֐ɿ੓࣏ എܠ ⼭⼝真⼀（2022）『ソーシャルメディア解体全書』、勁草書房 l ϑΣΠΫχϡʔεʹΑͬͯର৅ ΁ͷࢧ࣋Λ௿Լͤ͞ΔޮՌ l ಛʹऑ͍ࢧ࣋Λࢯ͍ͯͨਓʑʹ ର͠ϑΣΠΫχϡʔε͸ޮՌత ʹӨڹΛ༩͑Δ
9. 10 ϑΣΠΫχϡʔεΛආ͚Δࢦ਑ͷ࡞੒ എܠ https://www.weforum.org/agenda/2020/11/tackling-covid-19-misinformation-who/
10. 11 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥʢΞςϯγϣϯɾΤίϊϛʔʣ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
11. 12 ιʔγϟϧϝσΟΞ͸৘ใ֦ࢄ૷ஔ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε
12. 13 l ೔ຊͰ͸ϛυϧϝσΟΞͷଘࡏ͕ιʔγϟϧϝσΟΞͰͷϑΣΠ Ϋχϡʔε֦ࢄͷҰ໾Λ୲͍ͬͯΔͱ͍͏ࢦఠ https://www.soumu.go.jp/main_content/000749423.pdf ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε
13. 14 ϑΣΠΫχϡʔεͷ֦ࢄ͸଎ͨ͘͘͞Μ఻ΘΔ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε Vosoughi, Soroush, Deb Roy, and Sinan Aral. "The spread of true and false news online." science 359.6380 (2018): 1146-1151. https://www.soumu.go.jp/main_content/000630427.pdf l ಛʹ੓࣏ɺϏδωεɺઓ ૪ɺՊֶͳͲͷ࿩୊Ͱ֦ ࢄ͞Ε΍͍͢܏޲ l #PUͷޮՌ͸ҙ֎ͱখ͍͞
14. 15 ϑΣΠΫχϡʔεΛ֦ࢄ͢ΔϢʔβͱ͸ʁ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε Yamaguchi, Shinichi, and Tsukasa Tanihara. "Relationship between misinformation spreading behaviour and true/false judgments and literacy: an empirical analysis of COVID-19 vaccine and political misinformation in Japan." Global Knowledge, Memory and Communication (2023). ౰વ͕ͩɺͦͷχϡʔεΛʮਖ਼͍͠ʯͱ ৴ͨ͡ΓϝσΟΞɾϦςϥγʔ͕௿͍ਓ ͕֦ࢄ͠қ͍ ϑΣΠΫχϡʔεͷڞ༗ͷ͸ͷ ϢʔβʹΑΔ΋ͷͰɺߴྸऀ΍ӈدΓͷ ੓࣏ࢥ૝ͷϢʔβ͕ڞ༗͠΍͍͢ Grinberg, Nir, et al. "Fake news on Twitter during the 2016 US presidential election." Science 363.6425 (2019): 374-378.
15. 16 ϑΣΠΫχϡʔεΛݕग़͢ΔࢼΈ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε l ͲͷΑ͏ͳಛ௃Λ༻͍Δͷ͔ʁ l ༻͍ͨಛ௃Λ্ख͘׆༻Ͱ͖Δ͔ػցֶशϞσϧΛͲͷΑ͏ʹ ߏங͢Δͷ͔ʁ l ͲͷΑ͏ͳ෇ՃՁ஋ΛՃ͑Δͷ͔ʁʢ৴པੑͳͲʣ 機械学習モデル フェイクニュースの 可能性のある投稿に 関する情報 (特徴量) ⼊⼒ 出⼒ Fake or True
16. 17 ϑΣΠΫχϡʔεΛݕग़Ϟσϧͷྫ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε l ςΩετ৘ใ΍ڞ༗ͨ͠Ϣʔ βͷಛ௃Λ༻͍ͯɺϑΣΠΫ χϡʔε͔Ͳ͏͔Λ൑ఆ l ൑ఆʹ஫໨ͨ͠ՕॴΛՄࢹԽ ʢઆ໌ੑͷ෇༩ʣ Lu, Yi-Ju, and Cheng-Te Li. "GCAN: Graph-aware Co-Attention Networks for Explainable Fake News Detection on Social Media." Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020.
17. 18 ϑΣΠΫχϡʔεΛݕग़Ϟσϧͷ՝୊ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε l ೥ͷσʔλͰߏஙͨ͠ϑΣΠΫχϡʔεݕग़Ϟσϧ͕೥ ͷσʔλʹରԠͰ͖Δͷ͔ʁ
    EJBDISPOJD
    CJBTͷଘࡏ l ϞσϧΛຖ೥ߋ৽͠ଓ͚Δͱ͍͏ͷ͸େม l ߏங͞ΕΔػցֶशϞσϧͷ൑அ͸ਖ਼͍͠Θ͚Ͱ͸ͳ͍ l ਖ਼͍͠৘ใΛϑΣΠΫͱࢦఠ͢Δةݥੑʹ͍ͭͯ
    ِཅੑ΍ِӄੑ l ػցֶशϞσϧͷϒϥοΫϘοΫε໰୊ l ൑அͷࠜڌ΍࢓૊Έ͕Θ͔Βͳ͍ͱਓؒ͸ͦͷϞσϧΛ৴͖͡Εͳ͍ l େྔʹଘࡏ͢Δ౤ߘΛͲͷΑ͏ʹࡹ͔͘
18. 19 (VBSEJBOͷ׆༻ ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε   
       
     l (VBSEJBO
    ϑΣΠΫχϡʔεΛࢦఠ͢Δ Ϣʔβ l (VBSEJBOΛࣗಈతʹൃݟɺੜ੒͢Δ͜ ͱͰϑΝΫτνΣοΫίετͷ࡟ݮ l గਖ਼Ϣʔβ͕ެࣜΞΧ΢ϯτͳͲͰ͋Δ ͱޡใͷࢦఠΛ৴͡ΔՄೳੑ͕Ξοϓ Hannak, Aniko, et al. "Get back! you don’t know me like that: The social mediation of fact checking interventions in twitter conversations." Eighth International AAAI Conference on Weblogs and Social Media. 2014. Vo, Nguyen, and Kyumin Lee. "Learning from fact-checkers: analysis and generation of fact-checking language." Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2019. Seo, Haeseung, et al. "If You Have a Reliable Source, Say Something: Effects of Correction Comments on COVID-19 Misinformation." Proceedings of the International AAAI Conference on Web and Social Media. Vol. 16. 2022.
19. 20 ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε ϑΣΠΫχϡʔεͷ֦ࢄɿ೔ຊͷྫ https://www.asahi.com/articles/ASN2X6CXLN2XULFA03L.html https://www.asahi.com/articles/ASQ5D7X35Q5BULEI00R.html
20. 21 ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε 
    ʮ$07*%ʹΑͬͯτΠϨοτϖʔύʔ͕඼ബ͸ޡΓʯ ͱ͍͏౤ߘ͕ສҎ্ͷ35
21. 22 ྫ
    τΠϨοτϖʔύʔ૽ಈ
    JO
    $07*% ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε ೥݄೔͔Β೥݄೔ͷτΠϨοτϖʔύʔ͕ ؚΉ 
    UXFFUT
    
22. 23 ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε
23. 24 ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε 
    $07*%ʹΑΔτΠϨοτϖʔύʔ඼੾Ε౤ߘ
    35ҎԼ
    
    ࡟আࡁΈ
24. 25 ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε τΠϨοτϖʔύʔͷ඼੾Εͷใࠂ 21. より詳細な分析: ⿃海不⼆夫, 榊剛史, and 吉⽥光男. "ソーシャルメディアを⽤いた新型コロナ禍における感情変化の分析." ⼈⼯知能学会論⽂誌 35.4 (2020): F-K45_1.
25. 26 l ೔ຊͰ͸ϛυϧϝσΟΞͷଘࡏ͕ιʔγϟϧϝσΟΞͰͷϑΣΠ Ϋχϡʔε֦ࢄͷҰ໾Λ୲͍ͬͯΔͱ͍͏ࢦఠ https://www.soumu.go.jp/main_content/000749423.pdf ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε
26. 27 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥʢΞςϯγϣϯɾΤίϊϛʔʣ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
27. 28 ਓʑͷ࣋ͭೝ஌όΠΞε ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l ଟ͘ͷਓʑ͸ܦݧ΍௚ײ͔ Βޡͬͨ൑அΛͯ͠͠·͏ l ͜ΕΒͷόΠΞε͕ਓʑΛ ϑΣΠΫχϡʔεΛ৴͡͞ ͤΔޮՌΛ΋ͨΒ͢ҰํͰ ɺϑΣΠΫχϡʔεରࡦʹ ࢖͑ΔՄೳੑ΋ൿΊ͍ͯΔ https://en.wikipedia.org/wiki/Cognitive_bias
28. 29 όοΫϑΝΠΞʔޮՌ ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l ݸਓͷ৴೦ʹ߹Θͳ͍గਖ਼Λ༩͑ΒΕͨ ͱͯ͠΋ɺͦͷగਖ਼Λ৴ͣ͡ɺ͔͑ͬͯ గਖ਼ʹΑͬͯݸਓͷ৴೦ΛڧΊΔޮՌ l όοΫϑΝΠΞޮՌΛࣔ͢Ϣʔβ͸ɺෆ ฏෆຬ΍εϥϯάతͳදݱΛ༻͍Δ͜ͱ ͕ଟ͍ Nyhan, Brendan, and Jason Reifler. "When corrections fail: The persistence of political misperceptions." Political Behavior 32.2 (2010): 303-330. Jiang, Shan, and Christo Wilson. "Linguistic signals under misinformation and fact-checking: Evidence from user comments on social media." Proceedings of the ACM on Human-Computer Interaction 2.CSCW (2018): 1-23.
29. 30 "WBJMBCJMJUZ
    )FVSJTUJD ʢར༻ՄೳੑώϡʔϦεςΟοΫʣ ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l ࠷ۙ໨ʹͨ͠৘ใ΍සൟʹ઀৮͢Δ৘ใ ͕൑அʹӨڹΛ༩͑Δ͜ͱ l ϝσΟΞ΍4/4Ͱසൟʹ໨ʹ͢ΔϑΣΠ Ϋχϡʔε͸ɺޡͬͨ৘ใ͕ࣄ࣮ͱͯ͠ ೝࣝ͞Ε΍͍͢ Pennycook, Gordon, et al. "Shifting attention to accuracy can reduce misinformation online." Nature 592.7855 (2021): 590-595.
30. 31 ΫΠζΛ׆༻ͨ͠ϑΣΠΫχϡʔε΁ͷରࡦ ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l ϑΣΠΫχϡʔεΛݟ͔ͯΒగਖ਼͢ΔͷͰ ͸ͳ͘ɺʮڭҭʯͱͯ͠ϑΣΠΫχϡʔε ʹର͢ΔϫΫνϯΛ઀छ l ΫΠζܗࣜͰֶͿ͜ͱͰɺϑΣΠΫχϡʔ εʹରͯ͠ὃ͞Εʹ͘͘ͳΔޮՌΛൃش Roozenbeek, Jon, and Sander van der Linden. "Breaking Harmony Square: A game that “inoculates” against political misinformation." The Harvard Kennedy School Misinformation Review (2020).
31. 32 ಈըΛ׆༻ͨ͠ϑΣΠΫχϡʔε΁ͷରࡦ ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l ಈըʹΑΔϑΣΠΫχϡʔεରࡦ l ϑΣΠΫχϡʔεʹରͯ͠ڧ͘ͳΔʮ৺ ཧత༧๷઀छʯͷޮՌ͸ݟΒΕ͕ͨɺΫ Πζ΄Ͳͷେ͖͞Ͱ͸ͳ͍ l ҰํͰɺίετ͕҆͘ɺଟ͘ͷਓʑʹ ʮडಈతʯʹ઀छ͕Մೳ Roozenbeek, Jon, et al. "Psychological inoculation improves resilience against misinformation on social media." Science advances 8.34 (2022): eabo6254.
32. 33 φοδΛ׆༻ͨ͠ϑΣΠΫχϡʔε΁ͷରࡦ ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ l χϡʔεΛݟΔલʹ5SVF'BLFϥϕϧΛఏ ࣔ͢Δ΂͖͔ʁݟͨޙʹఏࣔ͢Δ΂͖͔ɺ ʹ͍ͭͯͲͷλΠϛϯάͰఏࣔ͢Δ͔Ͱ ਓʑͷϑΣΠΫχϡʔεʹର͢Δೝ஌͕ม Խ͢Δ͔Λݕূ l ݟग़͠ΛಡΜͩޙʹగਖ਼͢Δ͜ͱ͕Ұ൪ޮ Ռ͕ߴ͍͜ͱΛࣔ͢ Brashier, Nadia M., et al. "Timing matters when correcting fake news." Proceedings of the National Academy of Sciences 118.5 (2021): e2020043118.
33. 34 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥʢΞςϯγϣϯɾΤίϊϛʔʣ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
34. 35 ੜ੒"*ͷൃల ੜ੒"*ͱϑΣΠΫχϡʔε
35. 36 ੜ੒"*ʹΑΔ%FFQGBLFͷ࡞੒ ੜ੒"*ͱϑΣΠΫχϡʔε from: https://gulfbusiness.com/deepfakes-novel-trend-or-novel-threat/ l %FFQGBLF
    %FFQ
    -FBSOJOHͱ GBLFΛ૊Έ߹Θ͋Θͤͨ଄ޠ l Πϯλʔωοτ্Ͱ؍࡯͞Εͨ %FFQGBLFͷ਺͸૿Ճ܏޲
36. 37 'BLF/FXT
    (FOFSBUJPO ੜ੒"*ͱϑΣΠΫχϡʔε Brown, Tom, et al. "Language models are few-shot learners." Advances in neural information processing systems 33 (2020): 1877-1901. (15ʹΑΔχϡʔεੜ੒
37. 38 'BLF/FXT
    (FOFSBUJPO ੜ੒"*ͱϑΣΠΫχϡʔε https://news.yahoo.co.jp/expert/articles/846ed869620be2126f3cb749df6139d7282eb975 l Ұ໨Ͱ͸ݟ෼͚͕͔ͭͳ͍ը૾ ੜ੒ʹΑΔσϚ ྫɿz4IJ[VPLB
    EJTBTUFS
    GMPPEz IUUQTTUBCMFEJGGVTJPO XFCDPNQMBZHSPVOE
38. 39 ੜ੒"*ͷ੓࣏৘ใ΁ͷ׆༻ l ੓࣏ɾࣾձ໰୊ʹؔ͢Δ৘ใ Λ࡞੒͢ΔͨΊʹɺը૾ɾς Ωετͷੜ੒ϞσϧΛͷࠃ Ͱར༻͞Ε͍ͯΔ͜ͱΛใࠂ House, Freedom. The repressive power of artificial intelligence. Technical Report.{Freedom House}. https://freedomhouse. org/report/freedom-net/2023/repressive-power-artificial-intelligence, 2023. ੜ੒"*ͱϑΣΠΫχϡʔε
39. 40 ੜ੒"*ʹΑΔ࡞੒͞Εͨχϡʔεͷઆಘྗ l (15͸ਓؒΑΓ΋આಘྗͷ͋Δِ৘ใΛ࡞Γग़͢ l ਓؒʹ͸χϡʔε͕ੜ੒"*ʹΑΔ΋ͷ͔Ͳ͏͔൑அ Ͱ͖ͳ͍ Spitale, Giovanni, Nikola Biller-Andorno, and Federico Germani. "AI model GPT-3 (dis) informs us better than humans." Science Advances 9.26 (2023): eadh1850. ੜ੒"*ͱϑΣΠΫχϡʔε
40. 41 ϑΣΠΫχϡʔεݕग़ʹ--.͕࢖͑Δ͔ʁ Li, Guanghua, et al. "Re-Search for The Truth: Multi-round Retrieval-augmented Large Language Models are Strong Fake News Detectors." arXiv preprint arXiv:2403.09747 (2024). ੜ੒"*ͱϑΣΠΫχϡʔε l --.ͱݕࡧΛ૊Έ߹ΘͤΔ͜ͱͰχϡʔε͕'BLF͔Ͳ͏͔Λ൑ఆ
41. 42 %FFQGBLFݕग़ٕज़ͷൃల Reiss, Tal, Bar Cavia, and Yedid Hoshen. "Detecting deepfakes without seeing any." arXiv preprint arXiv:2311.01458 (2023). ੜ੒"*ͱϑΣΠΫχϡʔε https://www.i.u-tokyo.ac.jp/news/press/2022/202204262039.shtml
42. 43 ੜ੒"*࣌୅ͷϑΣΠΫχϡʔε ੜ੒"*ͱϑΣΠΫχϡʔε l ϑΣΠΫχϡʔεͷߴ౓ԽɾେྔԽ l ΠϯϑΥΧϦϓεͷ౸དྷʢʁʣ l ϩγΞɾ΢ΫϥΠφઓ૪͕ࣔ͢Α͏ʹɺ৘ใઓ͕ॏཁͳཱͪҐஔͱ ͳΔ l ϓϥοτϑΥʔϜࣄۀऀͷ໾ׂ͕ॏཁͱͳΔ
43. 44 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥͱ·ͱΊ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
44. 45 ΤίʔνΣϯόʔݱ৅ l 4/4্Ͱɺࣗ෼ͱࣅͨΑ͏ͳՁ஋؍΍ ߟ͑ํͷϢʔβʔΛϑΥϩʔ͢Δ͜ͱ Ͱɺಉ͡Α͏ͳχϡʔε΍৘ใ͹͔Γ ͕ྲྀ௨͢Δดͨ͡৘ใ؀ڥ l ෼ۃԽ΍ϑΣΠΫχϡʔεຮԆͷԹচ 4/4ΛऔΓר͘؀ڥ
45. 46 ೥݄ͷΞϝϦΧͷ4/4ۭؒ 4/4ΛऔΓר͘؀ڥ Nikolov, Dimitar, Alessandro Flammini, and Filippo Menczer. "Right and left, partisanship predicts (asymmetric) vulnerability to misinformation." arXiv preprint arXiv:2010.01462 (2020).
46. 47 ϑΟϧλʔόϒϧ l Ϣʔβͷݸਓ৘ใΛֶशͨ͠ΞϧΰϦζϜʹΑͬͯɺͦͷਓʹͱͬ ͯڵຯؔ৺͕͋Γͦ͏ͳ৘ใ͹͔Γ͕΍ͬͯ͘Δ৘ใ؀ڥ l (PPHMFݕࡧ΍4/4ͳͲ΄ͱΜͲͷαʔϏεʹඋ͑ΒΕ͍ͯΔਪન γεςϜͷӨڹʢύʔιφϥΠθʔγϣϯٕज़ͷෛͷଆ໘ʣ 4/4ΛऔΓר͘؀ڥ https://www.soumu.go.jp/main_content/000630427.pdf
47. 48 ΞςϯγϣϯɾΤίϊϛʔ l ϓϥοτϑΥʔϜͷ୆಄ʹΑ ΓΞςϯγϣϯࢢ৔ͷ୆಄ l ਓʑͷ࣌ؒΛୣ͍औΔڝ૪ۭ ؒͷܗ੒ l “Most people are unaware of how they spend their time and attention, their most crucial resources.” 4/4ΛऔΓר͘؀ڥ https://www.marketingmag.com.au/tech-data/why-attention-is-the-worlds-most-valuable- resource/
48. 49 ΞςϯγϣϯɾΤίϊϛʔ l Ξςϯγϣϯࢢ৔Ͱ͸ʮࢥ૝ͷڝ૪ʯ͔Βʮܹࢗͷڝ૪ʯ΁ l ϑΣΠΫχϡʔε͕૿͑Δཧ༝ͷͭ l 5SPMMͷ૿ՃͷཁҼʹ΋ l "*ͷ੒௕͸ਓʑͷ࠷దͳ"UUFOUJPO֫ಘํ๏Λ༧ଌ͢Δ l 5JL5PLͷॎεΫϩʔϧʹΑΔਪન l γεςϜʹΑͬͯਓʑͷڧ੍తͳʮ൓ࣹʯΛҾ͖ى͜͢͜ͱ͔Βɺ ʮनΘΕͷ௃ऩʢDBQUJWF
    BVEJFODFʣʯͱදݱ l ೔ຊͰ͸ిं಺ͷԻ੠޿ࠂ͕ਓ֨ݖ֐ʹ͋ͨΔ͔Ͱٞ࿦͕Ճ೤ͨ͜͠ͱ΋ 4/4ΛऔΓר͘؀ڥ
49. 50 ΞςϯγϣϯɾΤίϊϛʔ l 5JL5PLͷϨίϝϯυγ εςϜͷෆಁ໌ੑ΍ະ ੒೥΁ͷϦεΫ͔Βσ δλϧαʔϏε๏ͷٛ ຿ҧ൓ͷ͍͕͔͚ٙΒ ΕΔʢ&6ʣ 4/4ΛऔΓר͘؀ڥ https://www.asahi.com/articles/ASS2M7TT6S2MUHBI02B.html
50. 51 ࠓޙͷσδλϧ؀ڥ l ϓϥοτϑΥʔϜͷ໾ׂ͕·͢·͢ॏཁʹͳ͍ͬͯ͘ l Ͳ͜·Ͱن੍͠ɺͲ͜·Ͱදݱͷࣗ༝Λҡ࣋͢Δ؀ڥͱ͢Δͷ͔ʁ l ֎ࢿܥͷϓϥοτϑΥʔϜࣄۀऀ͕ଟ͍தͰɺϢʔβʹ೔ຊޠͰରԠͰ͖ Δମ੍Λ੔͑Δ͔ʁ l %FFQ'BLFɺϝλόʔεͳͲͷٕज़ͷֵ৽ʹ൐͍ɺੜ͡Δࣾձͷ ໰୊΋ෳࡶԽ l ৘ใϦςϥγʔڭҭͷ֦ॆ ͕ٸ຿ σδλϧ؀ڥͷະདྷ
51. 52 ࠓޙͷσδλϧ؀ڥ l ϑΝΫτνΣοΫ΍ίϛϡχςΟϊʔτͳͲͷऔΓ૊Έͷ֦ॆͷ ඞཁੑ l ೔ຊͰ͸ϑΝΫτνΣοΫػ͕ؔগͳ͍ʢϑΝΫτνΣοΫͷଟ༷ੑ͕อ ͨΕͳ͍ʣ l ϚεϝσΟΞ͕ͦͷ໾ׂΛ ୲͏͜ͱ͕ظ଴͞Ε͍ͯΔ σδλϧ؀ڥͷະདྷ ⼭⼝ほか（2022）「Innovation Nippon 2021 わが国における偽・誤情報の実 態の把握と 社会的対処の検討」 https://www.glocom.ac.jp/activities/project/7759
52. 53 ࠓޙͷσδλϧ؀ڥ ੍౓ɺγεςϜʢφοδʣɺϦςϥγʔͷͭΛಉ࣌ʹ ڧ͍ͯ͘͘͜͠ͱ͕ॏཁ l ೔ຊ͸ωοτ؀ڥ΍σδλϧίϯςϯπʹର͢Δ๏੍౓͕຀Վత l ϦςϥγʔڭҭɺϓϨόϯΩϯάͷऔΓ૊ΈΛૣΊʹऔΓ૊ΜͰ͍͘͜ͱ͕ॏ ཁ σδλϧ؀ڥͷະདྷ
53. 54 ൃද಺༰  എܠ  ιʔγϟϧϝσΟΞͱϑΣΠΫχϡʔε  ͳͥਓʑ͸ϑΣΠΫχϡʔεΛ৴ͯ͡͠·͏ͷ͔ʁ  ੜ੒"*ͱϑΣΠΫχϡʔε  4/4ΛऔΓר͘؀ڥͱ·ͱΊ  ࠷ۙͷऔΓ૊Έʢӄ๳࿦ͱಡॻʣ
54. 56 ಡॻཤྺ͔Βӄ๳࿦܏౗ϢʔβΛൃݟΛ໨ࢦ͢ ࠷ۙͷऔΓ૊Έ l ಡॻه࿥αʔϏεɿಡॻϝʔλʔ l ֤Ϣʔβ͕ʮ͍ͭʯɺʮԿͷຊʯΛͲͷΑ͏ͳײ૝Λ࣋ͬͯಡΜ͔ͩه࿥ l ӄ๳࿦ʹؔ͢ΔຊΛಡΜͩϢʔβͷཤྺ͕ݟΕΔ
55. 57 ಡॻཤྺσʔλͷ෼ੳ ࠷ۙͷऔΓ૊Έ l ܭࢉػՊֶతΞϓϩʔνΛద༻Մೳͳܗʹม׵ 埋め込み表現技術 ベクトル表現 0.1, 0.3, 0.4, …, 0.9, 0.2, 0.4 0.2, 0.1, 0.3, …, 0.1, 0.0, 0.2 0.3, 0.5, 0.1, …, 0.0, 0.3, 0.5 Book1 Book2 Book3 Book4 Book1 Book2 Book3
56. 58 ຊͷҐஔ͚ͮͷՄࢹԽ ࠷ۙͷऔΓ૊Έ
57. 59 ࠷ۙͷऔΓ૊Έ ॻ੶ͷຒΊࠐΈදݱ
58. 60 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ ωοτϫʔΫ ΫϥελϦϯά ࣌ؒܦա աڈͷಡॻཤྺ λʔήοτϢʔβ λʔήοτϢʔβ܈͕ӄ๳࿦ॻ੶ʹग़ձ͏·ͰʹಡΜͰ͖ͨຊ͔Βɼଞͷ Ϣʔβͱൺֱͯ͠ಛʹಡ·Ε͍ͯΔॻ੶ΫϥεΛݟ͚ͭΔ
59. 61 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ
60. 62 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ
61. 63 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ
62. 64 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ
63. 65 ࠷ۙͷऔΓ૊Έ 32
    ӄ๳࿦܏౗ಡऀʢλʔήοτϢʔβ܈ʣ͸ͲͷΑ͏ͳॻ੶Λ޷ΜͰಡΉ܏ ޲ʹ͋ΔͷͩΖ͏͔ʁ