{"id":9232,"date":"2025-10-12T20:54:42","date_gmt":"2025-10-12T20:54:42","guid":{"rendered":"https:\/\/pokecon.jp\/job\/?p=9232"},"modified":"2025-10-12T20:54:42","modified_gmt":"2025-10-12T20:54:42","slug":"ai%e3%81%aeux%e3%81%af%e3%80%8eguidelines-for-human-ai-interaction%e3%80%8f%e3%81%ab%e5%9f%ba%e3%81%a5%e3%81%84%e3%81%a6%e6%b1%ba%e3%82%81%e3%82%88%e3%81%86-aces-meet%e3%81%ae","status":"publish","type":"post","link":"https:\/\/pokecon.jp\/job\/9232\/","title":{"rendered":"AI\u306eUX\u306f\u300eGuidelines for Human-AI-Interaction\u300f\u306b\u57fa\u3065\u3044\u3066\u6c7a\u3081\u3088\u3046\u2014\u2014ACES Meet\u306e\u4e8b\u4f8b\u3092\u6dfb\u3048\u3066\u2014\u2014 – ACES \u30a8\u30f3\u30b8\u30cb\u30a2\u30d6\u30ed\u30b0"},"content":{"rendered":"\n<\/p>\n
\n

<\/span><\/p>\n

\u3053\u3093\u306b\u3061\u306f\u3002\u682a\u5f0f\u4f1a\u793eACES\u306eAI\u30a8\u30f3\u30b8\u30cb\u30a2\u306e\u963f\u4e45\u6fa4\u3067\u3059\u3002\u9577\u3089\u304f\u904b\u52d5\u3068\u306f\u7121\u7e01\u3067\u3057\u305f\u304c\u3001\u6700\u8fd1\u306f\u6751\u4e0a\u6625\u6a39<\/a>\u3055\u3093\u306e\u30a8\u30c3\u30bb\u30a4\u300e\u8077\u696d\u3068\u3057\u3066\u306e\u5c0f\u8aac\u5bb6\u300f\u306b\u5553\u767a\u3055\u308c\u3066\u3001\u4e00\u65e5\u4e00\u56de\u306e\u30e9\u30f3\u30cb\u30f3\u30b0\u3092\u65e5\u8ab2<\/a>\u306b\u3057\u3066\u3044\u307e\u3059\uff08\u5f7c\u306f\u6bce\u65e51\u6642\u9593\u7a0b\u5ea6\u8d70\u308b\u305d\u3046\u3067\u3059\uff09\u3002HP\u306f\u6d88\u8cbb\u3057\u307e\u3059\u304c\u3001\u6700\u5927HP\u306f\u4e0a\u304c\u3063\u3066\u3044\u308b\u6c17\u304c\u3057\u307e\u3059\u306d\u3002<\/p>\n

\u3055\u3066\u3053\u306e\u30d6\u30ed\u30b0\u3067\u306f\u3001AI\u3068UX\u306e\u95a2\u4fc2\u6027\u306b\u3064\u3044\u3066\u5f0a\u793e\u3067\u7279\u306b\u610f\u8b58\u3057\u3066\u3044\u308b\u30dd\u30a4\u30f3\u30c8\u3068\u3001\u5f0a\u793e\u306eSaaS<\/a>\u300eACES Meet\u300f<\/a>\u306e\u958b\u767a\u306b\u304a\u3051\u308b\u5b9f\u8df5\u5185\u5bb9\u306b\u3064\u3044\u3066\u7d39\u4ecb\u3057\u3066\u3044\u304d\u307e\u3059\u3002AI\u3068UX\u306e\u95a2\u4fc2\u6027\u306b\u3064\u3044\u3066\u306f\u3001\u65e5\u672c\u306eAI\u4f01\u696d\/\u958b\u767a\u8005\u306e\u4e2d\u3067\u5f37\u304f\u610f\u8b58\u3055\u308c\u3066\u3044\u308b\u3068\u8a8d\u8b58\u3057\u3066\u3044\u307e\u3059\u3002\u4f8b\u3048\u3070\u3001LayerX \u798f\u5cf6\u3055\u3093\u306e\u30d6\u30ed\u30b0\u306e\u4ee5\u4e0b\u306e\u8a00\u8449<\/p>\n

\n

AI\u3067\u4f53\u9a13\u3092\u69cb\u7bc9\u3059\u308b\u90e8\u5206\u3082\u3001AI\u306e\u7cbe\u5ea6\u306f100%\u306b\u306a\u3089\u306a\u3044\u3068\u3044\u3046\u524d\u63d0\u3067\u3001\u3044\u304b\u306b\u305d\u306e\u9593\u9055\u3044\u3092UX\u7684\u306b\u5438\u53ce\u3067\u304d\u308b\u304b\u3092\u8003\u3048\u308b\u5fc5\u8981\u304c\u3042\u308a\u307e\u3059\u3002<\/p>\n

\u300eAI-UX\u3068AX(AI Transformation)\u3068\u3044\u3046LayerX\u306e\u6311\u6226<\/a>\u300f\u3088\u308a\u5f15\u7528<\/p>\n<\/blockquote>\n

\u306f\u3001AI\u3092\u6a5f\u80fd\u306b\u843d\u3068\u3057\u8fbc\u3093\u3067\u3044\u304f\u969b\u306e\u30dd\u30a4\u30f3\u30c8\u3092\u7aef\u7684\u306b\u8868\u3057\u3066\u3044\u308b\u3068\u8003\u3048\u307e\u3059\u3002<\/p>\n

\u3055\u3066\u3001\u5b9f\u969b\u306b\u6a5f\u80fd\u958b\u767a\u3092\u3059\u308b\u4e0a\u3067\u300cAI\u306e\u9593\u9055\u3044\u3092UX\u7684\u306b\u5438\u53ce\u3059\u308b\u300d\u3068\u306f\u3001\u5177\u4f53\u7684\u306b\u3069\u3046\u3044\u3063\u305f\u3053\u3068\u3067\u3057\u3087\u3046\u304b\uff1f\u3053\u306e\u3088\u3046\u306aAI\u3068UX\u306e\u95a2\u4fc2\u6027\u306b\u3064\u3044\u3066\u8003\u3048\u308b\u6642\u3001\u5f0a\u793e\u3067\u306fMicrosoft<\/a> Research\u304c\u63d0\u5531\u3057\u305f\u300eGuidelines for Human-AI-Interaction\u300f<\/a>\uff08\u4ee5\u4e0b\u3001HAI\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a><\/strong>\uff09*1<\/a>\u3092\u53c2\u8003\u306b\u3059\u308b\u3053\u3068\u304c\u591a\u3044\u3067\u3059\u3002\u3053\u306e\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\u306f\u4eba\u3068AI\u304c\u5354\u50cd\u3059\u308bUX\u8a2d\u8a08\u306e\u30d9\u30b9\u30c8\u30d7\u30e9\u30af<\/a>\u30c6\u30a3\u30b9\u3092\u4f53\u7cfb\u7684\u306b\u6574\u7406\u3057\u305f\u3082\u306e\u3067\u3042\u308a\u3001\u5177\u4f53\u306e\u6a5f\u80fd\u958b\u767a\u306b\u304a\u3044\u3066\u5b88\u308b\u3079\u304d\u539f\u5247\u3084\u8003\u3048\u308b\u3079\u304d\u89b3\u70b9\u3092\u6559\u3048\u3066\u304f\u308c\u307e\u3059\u3002<\/p>\n

HAI\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\u306fChatGPT\u306e\u767b\u5834\u3088\u308a\u524d\u306e2019\u5e74\u306b\u767a\u8868\u3055\u308c\u305f\u3082\u306e\u3067\u3059\u304c\u3001\u305d\u306e\u30a8\u30c3\u30bb\u30f3\u30b9\u306f\u307e\u3060\u9306\u3073\u3066\u304a\u3089\u305a\u3001\u3080\u3057\u308dLLM\u6642\u4ee3\u306b\u3042\u3063\u3066\u305d\u306e\u91cd\u8981\u6027\u3092\u5897\u3057\u3066\u3044\u308b\u3082\u306e\u3068\u8003\u3048\u3066\u3044\u307e\u3059\u3002\u5f0a\u793e\u306eResearch Fellow\u3067\u3042\u308aCHI\u306e\u5c02\u9580\u5bb6\u3067\u3042\u308briku_arakawa\u3055\u3093\uff08@rikky0611<\/a>\uff09\u304c\u793e\u5185\u306b\u666e\u53ca\u3055\u305b\u3001AI\u3092\u6a5f\u80fd\u306b\u843d\u3068\u3057\u8fbc\u3080\u969b\u306eUX\u8a2d\u8a08\u306e\u6307\u91dd\u3068\u3057\u3066\u5f0a\u793e\u3067\u91cd\u5b9d\u3055\u308c\u3066\u304d\u307e\u3057\u305f\u3002<\/p>\n

Microsoft<\/a> Research \u304c 2019 \u5e74\u306b\u63d0\u5531\u3057\u305f\u300eGuidelines for Human-AI-Interaction\u300f\u306f\u3001\u4eba\u3068AI\u304c\u5354\u50cd\u3059\u308b\u969b\u306e\u30e6\u30fc\u30b6\u30fc\u4f53\u9a13\u8a2d\u8a08\u306b\u95a2\u3059\u308b18\u306e\u539f\u5247\uff08\u30c7\u30b6\u30a4\u30f3\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\uff09\u3092\u30014\u3064\u306e\u30d5\u30a7\u30fc\u30ba\u306b\u6cbf\u3063\u3066\u4f53\u7cfb\u7684\u306b\u6574\u7406\u3057\u305f\u3082\u306e\u3067\u3059\u3002\u3053\u306e\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\u306b\u6cbf\u3063\u305f\u6a5f\u80fd\u3092\u958b\u767a\u3059\u308b \/ UX\u3092\u63d0\u4f9b\u3059\u308b\u3053\u3068\u3067\u3001\u30e6\u30fc\u30b6\u30fc\u304cAI\u3092\u4fe1\u983c\u3057\u3001\u5b89\u5fc3\u3057\u3066\u6a5f\u80fd\u3092\u5229\u7528\u3059\u308b\u3088\u3046\u306b\u306a\u308b\u3068\u671f\u5f85\u3067\u304d\u307e\u3059\u3002<\/p>\n

\u305d\u308c\u3067\u306f\u3001\u3072\u3068\u307e\u305a\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\u3092\u3056\u3063\u3068\u773a\u3081\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u30d5\u30a7\u30fc\u30ba <\/th>\n No <\/th>\n \u539f\u5247\u5185\u5bb9\uff08\u65e5\u672c\u8a9e\u610f\u8a33\uff09 <\/th>\n \u82f1\u8a9e\u539f\u6587 <\/th>\n<\/tr>\n<\/thead>\n
Initially
\uff08\u5c0e\u5165\u6642\uff09<\/strong> <\/td>\n		 G1 <\/td>\n \u30b7\u30b9\u30c6\u30e0\u304c\u4f55\u3092\u3067\u304d\u308b\u304b\u3092\u660e\u793a\u3059\u308b <\/td>\n G1. Make clear what the system can do.<\/strong>Help the user understand what the AI system is capable of doing. <\/td>\n<\/tr>\n
<\/td>\n G2 <\/td>\n \u30b7\u30b9\u30c6\u30e0\u304c\u3069\u306e\u7a0b\u5ea6\u3046\u307e\u304f\u3067\u304d\u308b\u304b\u3092\u660e\u793a\u3059\u308b <\/td>\n G2. Make clear how well the system can do what it can do.<\/strong>Help the user understand how often the AI system may make mistakes. <\/td>\n<\/tr>\n
During Interaction
\uff08\u5229\u7528\u4e2d\uff09<\/strong> <\/td>\n		 G3 <\/td>\n \u6587\u8108\u306b\u5fdc\u3058\u3066\u30bf\u30a4\u30df\u30f3\u30b0\u3088\u304f\u30b5\u30fc\u30d3\u30b9\u3092\u63d0\u4f9b\u3059\u308b <\/td>\n G3. Time services based on context.<\/strong>Time when to act or interrupt based on the user\u2019s current task and environment. <\/td>\n<\/tr>\n
<\/td>\n G4 <\/td>\n \u30e6\u30fc\u30b6\u30fc\u306e\u72b6\u6cc1\u306b\u95a2\u9023\u3057\u305f\u60c5\u5831\u3092\u63d0\u793a\u3059\u308b <\/td>\n G4. Show contextually relevant information.<\/strong>Display information relevant to the user\u2019s current task and environment. <\/td>\n<\/tr>\n
<\/td>\n G5 <\/td>\n \u793e\u4f1a\u7684\u30fb\u6587\u5316\u7684\u898f\u7bc4\u306b\u6cbf\u3063\u305f\u4f53\u9a13\u3092\u63d0\u4f9b\u3059\u308b <\/td>\n G5. Match relevant social norms.<\/strong>Ensure the experience is delivered in a way that users would expect, given their social and cultural context. <\/td>\n<\/tr>\n
<\/td>\n G6 <\/td>\n \u8a00\u8a9e\u3084\u632f\u308b\u821e\u3044\u306b\u504f\u898b\u3092\u542b\u307e\u305b\u306a\u3044 <\/td>\n G6. Mitigate social biases.<\/strong>Ensure the AI system\u2019s language and behaviors do not reinforce undesirable and unfair stereotypes and biases. <\/td>\n<\/tr>\n
When Wrong
\uff08\u5931\u6557\u6642\uff09<\/strong> <\/td>\n		 G7 <\/td>\n \u30e6\u30fc\u30b6\u30fc\u304cAI\u3092\u547c\u3073\u51fa\u305b\u308b\u624b\u6bb5\u3092\u63d0\u4f9b\u3059\u308b <\/td>\n G7. Support efficient invocation.<\/strong>Make it easy to invoke<\/a> or request the AI system\u2019s services when needed. <\/td>\n<\/tr>\n
<\/td>\n G8 <\/td>\n \u4e0d\u8981\u306b\u306a\u3063\u305fAI\u6a5f\u80fd\u3092\u7c21\u5358\u306b\u7d42\u4e86\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b <\/td>\n G8. Support efficient dismissal.<\/strong>Make it easy to dismiss or ignore undesired AI system services. <\/td>\n<\/tr>\n
<\/td>\n G9 <\/td>\n \u8aa4\u3063\u305f\u3068\u304d\u306b\u30e6\u30fc\u30b6\u30fc\u304c\u4fee\u6b63\u30fb\u56de\u5fa9\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b <\/td>\n G9. Support efficient correction.<\/strong>Make it easy to edit, refine, or recover when the AI system is wrong. <\/td>\n<\/tr>\n
<\/td>\n G10 <\/td>\n \u30e6\u30fc\u30b6\u30fc\u306e\u610f\u56f3\u304c\u66d6\u6627\u306a\u3068\u304d\u306b\u7bc4\u56f2\u3092\u9650\u5b9a\u3057\u3066\u5bfe\u5fdc\u3059\u308b <\/td>\n G10. Scope services when in doubt.<\/strong>Engage in disambiguation or gracefully degrade the AI system\u2019s services when uncertain about a user\u2019s goals. <\/td>\n<\/tr>\n
<\/td>\n G11 <\/td>\n AI\u304c\u306a\u305c\u305d\u306e\u7d50\u679c\u306b\u306a\u3063\u305f\u304b\u3092\u8aac\u660e\u3059\u308b <\/td>\n G11. Make clear why the system did what it did.<\/strong>Enable the user to access<\/a> an explanation of why the AI system behaved as it did. <\/td>\n<\/tr>\n
Over Time
\uff08\u7d99\u7d9a\u5229\u7528\uff09<\/strong> <\/td>\n		 G12 <\/td>\n \u904e\u53bb\u306e\u30a4\u30f3\u30bf\u30e9\u30af<\/a>\u30b7\u30e7\u30f3\u3092\u8a18\u61b6\u30fb\u53c2\u7167\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b <\/td>\n G12. Remember recent interactions.<\/strong>Maintain short-term memory and allow the user to make efficient references to that memory. <\/td>\n<\/tr>\n
<\/td>\n G13 <\/td>\n \u30e6\u30fc\u30b6\u30fc\u304b\u3089\u5b66\u7fd2\u3057\u3001\u4f53\u9a13\u3092\u30d1\u30fc\u30bd\u30ca\u30e9\u30a4\u30ba\u3059\u308b <\/td>\n G13. Learn from user behavior.<\/strong>Personalize the user\u2019s experience by learning from their actions over time. <\/td>\n<\/tr>\n
<\/td>\n G14 <\/td>\n \u66f4\u65b0\u30fb\u9069\u5fdc\u306f\u6df7\u4e71\u3092\u6700\u5c0f\u9650\u306b\u6291\u3048\u308b <\/td>\n G14. Update and adapt cautiously.<\/strong>Limit disruptive changes when updating and adapting the AI system\u2019s behaviors. <\/td>\n<\/tr>\n
<\/td>\n G15 <\/td>\n \u7d30\u304b\u3044\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u3092\u5f97\u3089\u308c\u308b\u3088\u3046\u306b\u3059\u308b <\/td>\n G15. Encourage granular feedback.<\/strong>Enable the user to provide feedback indicating their preferences during regular interaction with the AI system. <\/td>\n<\/tr>\n
<\/td>\n G16 <\/td>\n \u30e6\u30fc\u30b6\u30fc\u884c\u52d5\u306e\u7d50\u679c\u3068\u305d\u306e\u5f71\u97ff\u3092\u4f1d\u3048\u308b <\/td>\n G16. Convey the consequences of user actions.<\/strong>Immediately update or convey how user actions will impact future behaviors of the AI system. <\/td>\n<\/tr>\n
<\/td>\n G17 <\/td>\n AI\u306e\u76e3\u8996\u3084\u6319\u52d5\u3092\u30e6\u30fc\u30b6\u30fc\u304c\u5236\u5fa1\u3067\u304d\u308b\u3088\u3046\u306b\u3059\u308b <\/td>\n G17. Provide global controls.<\/strong>Allow the user to globally customize what the AI system monitors and how it behaves. <\/td>\n<\/tr>\n
<\/td>\n G18 <\/td>\n \u30b7\u30b9\u30c6\u30e0\u306e\u66f4\u65b0\u3084\u5909\u66f4\u3092\u901a\u77e5\u3059\u308b <\/td>\n G18. Notify users about changes.<\/strong>Inform the user when the AI system adds or updates its capabilities. <\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n

\u3053\u3061\u3089\u3092\u898b\u3066\u4f55\u304b\u67d3\u307f\u5165\u308b\u3082\u306e\u306f\u3042\u308a\u307e\u3057\u305f\u3067\u3057\u3087\u3046\u304b\u3002\u300c\u3046\u3093\u3046\u3093\u300d\u3068\u306a\u3063\u3066\u3044\u308b\u65b9\u3001\u3055\u3059\u304c\u3067\u3059\u3002\u5b9f\u52d9\u3067\u304b\u306a\u308aAI\u3092\u4f7f\u3063\u3066\u304a\u3089\u308c\u308b\u3053\u3068\u3067\u3057\u3087\u3046\u3002\u9006\u306b\u30a4\u30e1\u30fc\u30b8\u304c\u3064\u304b\u306a\u304b\u3063\u305f\u65b9\u3001\u3053\u306e\u6b21\u306e\u7ae0\u3067\u5f0a\u793e\u4e8b\u4f8b\u3092\u30b1\u30fc\u30b9\u30b9\u30bf\u30c7\u30a3<\/a>\u3068\u3057\u3066\u3054\u7d39\u4ecb\u3059\u308b\u306e\u3067\u3001\u305c\u3072\u305d\u3061\u3089\u3082\u3054\u89a7\u306b\u306a\u3063\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n

\u3055\u3066\u3001\u79c1\u3068\u3057\u3066\u306f\u4e0a\u8a18\u306e\u30ac\u30a4\u30c9\u30e9\u30a4\u30f3<\/a>\u306e\u7279\u5fb4\u3092\u300c\u4eba\u9593\u4e2d\u5fc3\u300d\u306e\u8996\u70b9<\/strong>\u3068\u8003\u3048\u3066\u3044\u307e\u3059\u3002\u4eba\u9593\u4e2d\u5fc3\u3068\u306f\u3064\u307e\u308a\u3001AI\u3092\u5358\u306a\u308b\u81ea\u52d5\u5316\u30c4\u30fc\u30eb\u3068\u3057\u3066\u3067\u306f\u306a\u304f\u3001\u4eba\u9593\u306e\u610f\u601d\u6c7a\u5b9a\u3084\u884c\u52d5\u3092\u88dc\u52a9\u3059\u308b\u30d1\u30fc\u30c8\u30ca\u30fc<\/strong>\u3068\u3057\u3066\u4f4d\u7f6e\u3065\u3051\u3066\u3044\u307e\u3059\u3002\u4f8b\u3048\u3070\u3001<\/p>\n