公開設計指引Disclosure design guidelines

瞭解如何在您的語音體驗綜合本質上透明化,以建立及維護與客戶的信任。Learn how to build and maintain trust with customers by being transparent about the synthetic nature of your voice experience.

什麼是洩漏?What is disclosure?

洩漏是一種可讓人們知道他們'重新互動或接聽合成方式產生之語音的方法。Disclosure is a means of letting people know they're interacting with or listening to a voice that is synthetically generated.

為什麼需要洩漏?Why is disclosure necessary?

公開電腦產生的語音之綜合來源的需求相當新。The need to disclose the synthetic origins of a computer-generated voice is relatively new. 在過去,電腦產生的語音顯然很明顯,因為沒有人會犯真正的聲音。In the past, computer-generated voices were obviously that—no one would ever mistake them for a real person. 不過,每一天都會改善綜合語音的真實性,並使其與人類語音的區別變得越來越不容易。Every day, however, the realism of synthetic voices improves, and they become increasingly indistinguishable from human voices.


設計綜合語音體驗時,請記住下列原則:These are the principles to keep in mind when designing synthetic voice experiences:

強化信任Reinforce trust
設計的目的是要讓 Turing 測試失敗,而不會降低經驗。Design with the intention to fail the Turing Test without degrading the experience. 讓使用者知道他們正在與綜合語音互動,同時讓他們能夠順暢地與經驗互動。Let the users in on the fact that they're interacting with a synthetic voice while allowing them to engage seamlessly with the experience.

適應使用內容Adapt to context of use
瞭解您的使用者將如何與綜合語音互動,以在正確的時機提供正確的洩漏類型。Understand when, where, and how your users will interact with the synthetic voice to provide the right type of disclosure at the right time.

設定明確的預期Set clear expectations
可讓使用者輕鬆地探索和瞭解代理程式的功能。Allow users to easily discover and understand the capabilities of the agent. 提供機會,讓您在提出要求時深入瞭解綜合語音技術。Offer opportunities to learn more about synthetic voice technology upon request.

接納失敗Embrace failure
使用失敗的時間來強化代理程式的功能。Use moments of failure to reinforce the capabilities of the agent.

如何使用本指南How to use this guide

本指南可協助您判斷哪些洩漏模式最適合您的綜合語音體驗。This guide helps you determine which disclosure patterns are best fit for your synthetic voice experience. 接著,我們會提供如何及何時使用它們的範例。We then offer examples of how and when to use them. 這些模式的每一種都是設計用來最大化使用者對綜合語音的透明度,同時又不會受到人為中心的設計。Each of these patterns is designed to maximize transparency with users about synthetic speech while staying true to human-centered design.

考慮到大量的語音體驗設計指引,我們特別著重于:Considering the vast body of design guidance on voice experiences, we focus here specifically on:

  1. 洩漏評估:判斷您的綜合語音體驗建議之洩漏類型的流程Disclosure assessment: A process to determine the type of disclosure recommended for your synthetic voice experience

  2. 如何公開:可套用至綜合語音體驗的公開模式範例How to disclose: Examples of disclosure patterns that can be applied to your synthetic voice experience

  3. 何時公開:在整個使用者旅程中公開的最佳時刻When to disclose: Optimal moments to disclose throughout the user journey

洩漏評量Disclosure assessment

請考慮您的使用者' 有關互動的期望,以及他們會遇到語音的內容。Consider your users' expectations about an interaction and the context in which they will experience the voice. 如果內容清楚指出綜合聲音是"說話的,"洩露可能是最小、瞬間,甚至不需要。If the context makes it clear that a synthetic voice is "speaking," disclosure may be minimal, momentary, or even unnecessary. 影響洩漏的主要內容類型包括「角色類型」、「案例類型」和「暴露程度」。The main types of context that affect disclosure include persona type, scenario type, and level of exposure. 它也有助於考慮誰可能正在接聽。It also helps to consider who might be listening.

瞭解內容Understand context

使用此工作表來判斷綜合語音體驗的內容。Use this worksheet to determine the context of your synthetic voice experience. 您會在下一個步驟中套用此功能,以判斷您的洩漏等級。You'll apply this in next step where you'll determine your disclosure level.

使用內容Context of use 潛在風險 & 挑戰Potential Risks & Challenges
1. 角色類型1. Persona type 如果符合下列任何一項,您的角色就會放在「人類贊角色」類別之下:If any of the following apply, your persona fits under the 'Human-like Persona' category:

  • 「角色」是一種真實的人,不論是虛構的標記法。Persona embodies a real human whether it's a fictitious representation or not. (例如,相片或電腦產生的真實人呈現)(e.g., photograph or a computer-generated rendering of a real person)

  • 綜合語音是以廣為辨識之真實人員的語音為基礎(例如名人、政治圖)The synthetic voice is based on the voice of a widely recognizable real person (e.g., celebrity, political figure)
讓您的角色擁有更相似的標記法,使用者越可能將它與真實的人建立關聯,或讓他們相信內容是由真實人員說出,而不是電腦產生的。The more human-like representations you give your persona, the more likely a user will associate it with a real person, or cause them to believe that the content is spoken by a real person rather than computer-generated.
2. 案例類型2. Scenario type 如果下列任何一項適用,您的語音體驗會符合 [敏感性] 類別:If any of the following apply, your voice experience fits under the 'Sensitive' category:

  • 取得或顯示使用者的個人資訊Obtains or displays personal information from the user

  • 廣播時間敏感性新聞/資訊(例如緊急警示)Broadcasts time sensitive news/information (e.g., emergency alert)

  • 旨在協助真實人員彼此通訊(例如,讀取個人電子郵件/文字)Aims to help real people communicate with each other (e.g., reads personal emails/texts)

  • 提供醫療/健全狀況協助Provides medical/health assistance
當主題與機密、個人或緊急事項相關時,使用綜合語音可能不會對使用它的人員感到適當或值得信任。The use of synthetic voice may not feel appropriate or trustworthy to the people using it when topics are related to sensitive, personal, or urgent matters. 他們可能也會預期與實際人一樣,也會有相同層級的理解和內容感知。They may also expect the same level of empathy and contextual awareness as a real human.
3. 公開程度3. Level of exposure 您的語音體驗最可能符合「高」類別,如果:Your voice experience most likely fits under the 'High' category if:

  • 使用者會聽到或長時間內的綜合語音,或經常與之互動The user will hear or interact with the synthetic voice frequently or for a long duration of time
在建立長期關聯性時,透明且與使用者建立信任的重要性更高。The importance of being transparent and building trust with users is even higher when establishing long-term relationships.

判斷洩漏層級Determine disclosure level

使用下列圖表來判斷您的綜合語音體驗是否需要根據您的使用內容來進行高或低洩漏。Use the following diagram to determine whether your synthetic voice experience requires high or low disclosure based on your context of use.


參考文件Reference docs

後續步驟Next steps