This notebook is an exercise in the AI Ethics course. You can reference the tutorial at this link.

---

In the tutorial, you learned about human-centered design (HCD) and became familiar with six general steps to apply it to AI systems. In this exercise, you will identify and address design issues in six interesting AI use cases.

在本教程中，您了解了以人为本的设计 (HCD)，并熟悉了将其应用于 AI 系统的六个一般步骤。在本练习中，您将识别并解决六个有趣的 AI 用例中的设计问题。

Introduction

简介

Begin by running the next code cell.

首先运行下一个代码单元。

• Clicking inside the code cell.
• 单击代码单元内部。
• Click on the triangle (in the shape of a "Play button") that appears to the left of the code cell.
• 单击代码单元左侧出现的三角形（形状为“播放按钮”）。
• If your code is run successfully, you will see Setup Complete as output below the cell.
• 如果您的代码成功运行，您将在单元下方看到“设置完成”作为输出。

# Set up feedback system
from learntools.core import binder
binder.bind(globals())
from learntools.ethics.ex2 import *
print("Setup Complete")

Setup Complete

1) Reducing plastic waste

1) 减少塑料垃圾

A Cambodian organization wants to help reduce the significant amounts of plastic waste that pollute the Mekong River System. Which of the following would be an appropriate way to start? (Your answer might use more than one option.)

柬埔寨的一个组织希望帮助减少污染湄公河系统的大量塑料垃圾。以下哪项是合适的开始方式？（您的答案可能使用多个选项。）

• Watch the people currently addressing the problem as they navigate existing tools and processes.
• 观察目前正在解决问题的人如何利用现有工具和流程。
• Conduct individual interviews with the people currently addressing the problem.
• 对目前正在解决问题的人进行个人访谈。
• Assemble focus groups that consist of people currently addressing the problem.
• 召集由目前正在解决问题的人组成的焦点小组。

After you have answered the question, view the official solution by running the code cell below.

回答完问题后，通过运行下面的代码单元查看官方解决方案。

# Check your answer (Run this code cell to receive credit!)
q_1.check()

Solution: These are all good ways to start!

这些都是很好的开始方式！

2) Detecting breast cancer

2) 检测乳腺癌

Pathologists try to detect breast cancer by examining cells on tissue slides under microscopes. This tiring and repetitive work requires an expert eye. Your team wants to create a technology solution that helps pathologists with this task in real-time, using a camera. However, due to the complexity of the work, your team has not found rule-based systems to be capable of adding value to the review of images.

病理学家试图通过在显微镜下检查组织载玻片上的细胞来检测乳腺癌。这项累人且重复的工作需要专家的眼光。您的团队希望创建一种技术解决方案，使用摄像头实时帮助病理学家完成这项任务。但是，由于工作的复杂性，您的团队尚未发现基于规则的系统能够为图像审查增加价值。

Would AI add value to a potential solution? Why or why not?

人工智能会为潜在解决方案增加价值吗？为什么或为什么不？

# Check your answer (Run this code cell to receive credit!)
q_2.check()

Solution: Yes, it would. People would generally agree that the goal is desirable, especially since the AI system will be working with pathologists rather than in their place. AI can help people with repetitive tasks and AI systems have proven effective in similar medical image recognition use cases. That said, it is important to follow current industry best practices and to be thorough in the rest of the design process, including in analyzing harms and in considering how medical practitioners will actually interact with the product in a medical setting.

是的，确实如此。人们普遍认为这个目标是可取的，尤其是因为人工智能系统将与病理学家合作，而不是代替他们工作。人工智能可以帮助人们完成重复性任务，而且人工智能系统在类似的医学图像识别用例中已被证明是有效的。话虽如此，重要的是遵循当前行业最佳实践，并在设计过程的其余部分做到彻底，包括分析危害和考虑医疗从业者在医疗环境中如何与产品进行实际互动。

3) Flagging suspicious activity

3) 标记可疑活动

A bank is using AI to flag suspicious international money transfers for potential money laundering, anti-terrorist financing or sanctions concerns. Though the system has proven more effective than the bank’s current processes, it still frequently flags legitimate transactions for review.

一家银行正在使用人工智能标记可疑的国际汇款，以防潜在的洗钱、反恐融资或制裁问题。尽管该系统已被证明比银行目前的流程更有效，但它仍然经常标记合法交易以供审查。

What are some potential harms that the system could cause, and how can the bank reduce the impacts of these potential harms?

该系统可能造成哪些潜在危害，银行如何减少这些潜在危害的影响？

# Check your answer (Run this code cell to receive credit!)
q_3.check()

Solution: One potential harm is that the AI system could be biased against certain groups, flagging, delaying or denying their legitimate transactions at higher rates than those of other groups. The bank can reduce these harms by selecting data carefully, identifying and mitigating potential bias (see Lessons 3 and 4), not operationalizing the system until potential bias is addressed and ensuring appropriate and continuous human oversight of the system once it is operational.

一个潜在的危害是，人工智能系统可能会对某些群体产生偏见，比其他群体更频繁地标记、延迟或拒绝他们的合法交易。银行可以通过谨慎选择数据、识别和减轻潜在偏见（参见第 3 和第 4 课）、在解决潜在偏见之前不投入使用系统以及确保在系统投入使用后对系统进行适当且持续的人工监督来减少这些危害。

4) Prototyping a chatbot

4) 制作聊天机器人原型

During an ongoing pandemic outbreak, a country’s public health agency is facing a large volume of phone calls and e-mails from people looking for health information. The agency has determined that an AI-powered interactive chatbot that answers pandemic-related questions would help people get the specific information they want quickly, while reducing the burden on the agency’s employees. How should the agency start prototyping the chatbot?

在持续的疫情爆发期间，一个国家的公共卫生机构面临着大量寻求健康信息的人们打来的电话和发来的电子邮件。该机构已确定，一个能够回答疫情相关问题的人工智能驱动的交互式聊天机器人将帮助人们快速获得他们想要的具体信息，同时减轻该机构员工的负担。该机构应如何开始制作聊天机器人原型？

• Build out the AI solution to the best of its ability before testing it with a diverse group of potential users.
• 在与多样化的潜在用户群体进行测试之前，尽最大努力构建人工智能解决方案。
• Build a non-AI prototype quickly and start testing it with a diverse group of potential users.
• 快速构建非人工智能原型，并开始与多样化的潜在用户群体进行测试。

# Check your answer (Run this code cell to receive credit!)
q_4.check()

Solution: The correct answer is: Build a non-AI prototype quickly and start testing it with a diverse group of potential users. Iterating on a non-AI prototype is easier, faster and less expensive than iterating on an AI prototype. Iterating on a non-AI prototype also provides early information on user expectations, interactions and needs. This information should inform the eventual design of AI prototypes.

正确答案是：快速构建非 AI 原型并开始与多样化的潜在用户群一起测试。非 AI 原型的迭代比 AI 原型的迭代更容易、更快且成本更低。非 AI 原型的迭代还可以提供有关用户期望、交互和需求的早期信息。这些信息应该为 AI 原型的最终设计提供参考。

5) Detecting misinformation

5) 检测错误信息

A social media platform is planning to deploy a new AI system to flag and remove social media messages containing misinformation. Though the system has proven effective in tests, it sometimes flags non-objectionable content as misinformation.

社交媒体平台计划部署一个新的人工智能系统来标记和删除包含错误信息的社交媒体消息。尽管该系统在测试中被证明是有效的，但它有时会将无异议的内容标记为错误信息。

What are some ways in which the social media platform could allow someone whose message has been flagged to contest the misinformation designation?

社交媒体平台可以通过哪些方式允许被标记消息的人对错误信息的认定提出异议？

# Check your answer (Run this code cell to receive credit!)
q_5.check()

Solution: The social media company should ask customers how they would want to challenge a determination. It could be by easily accessing a challenge form on which a user can describe why their message does not contain misinformation, requesting further review by a human reviewer, requesting an explanation of why the content was flagged or a combination of these and other means.

社交媒体公司应该询问客户他们希望如何质疑决定。可以采取以下方式：轻松访问质疑表单，用户可以在表单上描述其信息不包含错误信息的原因、请求人工审核人员进一步审核、请求解释内容被标记的原因，或者上述方式和其他方式的组合。

6) Improving autonomous vehicles

6) 改进自动驾驶汽车

What are some of the ways to improve the safety of autonomous vehicles? (You might pick more than one option.)

有哪些方法可以提高自动驾驶汽车的安全性？（您可以选择多个选项。）

• Incorporate the safety features of regular vehicles.
• 融入普通车辆的安全功能。
• Test the system in a variety of environments.
• 在各种环境中测试系统。
• Hire an internal ‘red team’ to play the role of bad actors seeking to manipulate the autonomous driving system. Strengthen the system against the team’s attacks on an ongoing basis.
• 聘请内部“红队”扮演试图操纵自动驾驶系统的坏人。持续加强系统抵御团队攻击的能力。

# Check your answer (Run this code cell to receive credit!)
q_6.check()

Solution: All of these are great ways to improve safety.

所有这些都是提高安全性的好方法。

Keep going

继续前进

Continue to identify bias in machine learning applications.

继续在机器学习应用中识别偏差。

---