Skip to main content
SearchLoginLogin or Signup

Navigating Biases in AI Algorithms

This course explores algorithmic bias, teaching learners strategies for critically evaluating and ethically using generative AI tools.

Published onOct 10, 2024
Navigating Biases in AI Algorithms
·

Summary

The rise in popularity of new AI tools like ChatGPT and AI-powered search tools has led to instances of misuse, such as the spread of false information on social media and the creation of inappropriate AI-generated images of students. Instead of attempting to prohibit these tools, schools should enhance their digital literacy programs by including modules on AI literacy to teach students how to use them responsibly. To incorporate these modules effectively, it is essential to provide information professionals, faculty partners, or educators with the necessary skills and knowledge to recognize, critically assess, and address biases present in online information on various platforms.

This course will explore biases inherent in AI algorithms and their use on online platforms like digital archives, social media, search engines, and other AI-driven services. It will encourage critical thinking about the credibility of information, the transparency of curation processes, the accuracy of its representations, and the complexities related to information authority, formats, and editorial oversight. Additionally, appropriate strategies will be taught to help individuals be mindful of and evaluate these biases present in each type of media.

Author

Seul Lee, Ph.D. Candidate, Information Studies, UCLA

Learning Outcomes

Upon satisfactory completion of this course, we will achieve the following learning outcomes:

  • Gain a comprehensive overview of the field of critical internet studies, empowering us to critically analyze and evaluate online information on various platforms and AI-driven tools.

  • Enhance their abilities in both technical and critical analysis of information sources.

  • Develop a critical understanding of the distinctive nature of different online platforms and tools, including their goals, affordances, practices, and preferences, and articulate relevant issues.

  • Acquire an understanding of how various socio-cultural contexts shape technological landscapes, leading to heightened awareness of the ethical implications embedded in online platforms.

  • Reflect more critically on the nature of online platforms and tools, fostering a thoughtful approach to their use and impact.

  • Utilize various online platforms and tools effectively and safely for everyday information seeking, incorporating privacy protection measures and the ability to identify reliable sources.

Audience

This course is designed to assist information professionals, faculty partners, and educators who want to incorporate critical digital literacy concepts into their teaching as well as college students who make use of different online platforms, including commercial search engines, social media platforms, and AI-driven tools for their everyday information seeking, to better understand how the values, business strategies, profit motives, and future plans of various stakeholders shape these online information sources.

Curricular Context

This course is tailored for beginner-level information professionals, faculty partners, educators, and librarians who wish to integrate critical digital literacy principles into their instruction, along with students and researchers, making it accessible to individuals with varying backgrounds and levels of expertise. No prior knowledge is required, allowing participants to embark on their learning journey with a fresh perspective and receptive attitude.

The course is designed to be conducted in a flexible format, allowing for both fully virtual delivery and in-person sessions with minimal adaptations. The course components include live webinars, online forums, group discussions, presentations, and hands-on workshops.

During the hands-on workshops, participants will be asked to choose two primary tools from a list that includes Microsoft Excel, Python, RStudio, GitHub, Tableau, Voyant, Gephi, p5.js, Cytoscape, Canva, or other data analysis and visualization tools for their own projects. Participants can suggest alternative tools such as Google Sheets for their projects. Participants should consider system requirements, as some tools may need to be downloaded and installed. They will then present their final projects that analyze their chosen online platform or tool at the end of the class. This requirement aims to provide students with hands-on experience and the opportunity to apply the concepts and techniques learned throughout the course in a practical setting. To enhance the workshop experience with tools like Microsoft Excel, Python, RStudio, GitHub, Tableau, etc., one to two lab sessions should be added. In addition to the primary instructor, one or two guest lecturers can be invited to share their projects and insights, providing students with diverse perspectives on the subject matter.

In consideration of accessibility for our virtual course, closed captions will be provided for all live webinars and recorded sessions. Participants are encouraged to reach out with any additional accessibility needs or accommodations they may require to fully engage with the course content.

Preparation

  • Instructors and participants need access to laptops or desktop computers with a reliable internet connection. For clearer communication, webcams and headsets/headphones with microphones are required.

  • For class activities, instructors and participants need access to the following software applications:  Microsoft Excel, Python, RStudio, GitHub, Tableau, Voyant, Gephi, p5.js, Cytoscape, and Canva. Some applications may require installation and user accounts for access, and they may have specific system requirements.

  • Before the course, participants will be required to complete any pre-reading assignments or watch introductory videos provided by the instructor. If necessary, instructors may assign preparatory tasks, such as installing required software applications and setting up user accounts.

For Instructors

  • In preparation for teaching this course, instructors should review and become familiar with all course materials, including presentations, hands-on workshops, and discussion topics.

  • Instructors are also encouraged to read and practice the tutorials available on the website (https://khistory.mobirisesite.com/) before teaching. These tutorials provide comprehensive guides on how to use the materials and tools effectively, ensuring instructors are well-prepared to support participants in their learning journey.

  • It is essential to ensure technical proficiency with the selected tools such as Microsoft Excel, Python, RStudio, GitHub, Tableau, Voyant, Gephi, p5.js, Cytoscape, Canva, and other relevant data analysis and visualization tools. Instructors should also understand and implement accessibility features like closed captions for webinars and recordings to accommodate diverse learners.

  • They should verify all provided links to ensure accuracy and accessibility for participants. Familiarizing with available support resources, including troubleshooting guides or assistance for tool-specific issues, is crucial for effective course delivery.

  • Instructors should be prepared to adapt the course based on participants' needs, considering varying levels of technical expertise and accessibility requirements.

  • Developing engagement strategies for online forums, group discussions, and workshops will help foster active participation and learning.

  • It is also important to establish methods for gathering feedback and assessing participant progress to ensure effective learning outcomes. Instructors can set up communication channels with participants to facilitate questions, support, and feedback throughout the course duration.

Materials

Most of the course materials listed as required readings can be found in the Lesson Outline below.

Textbooks:

There is one required textbook for this course:

Barocas, Solon, Moritz Hardt, and Arvind Narayanan. Fairness and Machine Learning: Limitations and Opportunities. The MIT Press, 2023.

Participants are not obligated to purchase it. Rather, a complimentary PDF version of the full text can be accessed and downloaded from the website: https://fairmlbook.org.

Programming language and software:

In our lab section, participants can learn about several programming languages and software packages and use them to complete several assignments. Participants are required to choose two primary tools to present their final projects:  Microsoft Excel, Python, RStudio, GitHub, Tableau, Voyant, Gephi, p5.js, Cytoscape, Canva, or other data analysis and visualization tools.

Supplementary Materials

Lesson Outline

Note: For a detailed overview of the course curriculum, see Lee-Weekly-Syllabus.pdf under Supplementary Materials

Participants will engage actively in discussions about the current challenges encountered by the profession and related areas, and improve their readiness to tackle these issues. While this course aims to equip learners with the skills needed to navigate the complex information landscape, whether the agenda is to sway public opinion, influence elections, or profit from misinformation, as well as to support accessibility issues, it will also address cultural and regional differences in information biases to foster a global understanding of the digital world and teach participants how to responsibly share information and contribute to a more transparent and reliable online environment. This course focuses on helping participants develop the skills to be knowledgeable and accountable digital citizens who can effectively navigate the challenges of dealing with biased information in online settings.

Furthermore, to enhance the workshop experience effectively utilizing various tools such as Microsoft Excel, Python, RStudio, GitHub, Tableau, Voyant, Gephi, p5.js, Cytoscape, Canva, and others, there will be one or two lab sessions. While having prior knowledge in data analysis or visualization would be advantageous for participants, as it can help reduce the learning curve and make the workshop content more accessible to individuals with diverse backgrounds and experiences, participants are expected to engage in the sessions and seek assistance during the lab time.

Sample Lesson Plan: Week 2 (Data-Driven Methods)

Context

This is the first lesson of the ‘Data-Driven Methods’ unit.

  • Participants will gain an understanding of data-driven methods through workshops that highlight data-driven projects in various fields.

  • During this session, participants will engage in a short survey to indicate which online platforms/services/tools they wish to explore further for a project. The instructor will then assess these responses to determine whether to retain or modify the current platforms or tools.

  • The core lesson should take around 1.5-2 hours. The schedule is adaptable to accommodate various schedules.

Activities

  • Attendance and recap of assignment (10 mins)

    • Welcome participants to their second lesson.

  • Introductory discussion (15 mins)

    • Explain that today we will be delving into the topic of data-driven research.

  • Introduce why we are going to learn about data-driven research today.

  • Ask: What is data-driven research?

    • Then ask: What are the components of data-driven research?

  • Participants share their ideas with the class. From their responses, compile a list outlining the elements of data-driven research.

  • Exemplary case tutorial (30 mins)

    • Explore the tutorial website to learn about step-by-step processes for data-driven research. Learn how to begin by formulating appropriate research questions, and finding the right data sources, tools, and research techniques. Additionally, learn how to clean and reorganize datasets for their own research purposes.

  • Break (10 mins)

  • Exemplary case tutorial w/Infographic slides (30 mins)

    • Briefly explore various data analysis tools and how they can be utilized. Participants can choose the tools that best suit their research needs after reviewing this overview.

  • Lecture (10 mins)

    • Briefly summarize the components of data-driven research and outline the necessary steps to conduct such research. Inquire if there are any additional queries from the participants.

  • A short survey (10 mins)

    • Share the Google Forms link for a short online survey to determine which online platforms, services, or tools should be explored more deeply for a project, along with the rationale for why they are considered important. (15 mins)

Teaching methods

The course spans a duration of 10-11 weeks, during which each session comprises a 1.5-2 hour lecture followed by a lab/workshop segment lasting between 30 minutes to 1 hour. However, the length of the course can be modified depending on the instructor’s availability and goals, as well as the level of learners.

The lecture and lab session, lasting 2.5-3 hours, will involve discussing articles, engaging in critical self-reflection on the course material, and presenting on selected platforms or tools. Each week the instructor will give a lecture and lead a discussion, supported by PowerPoint slides, handouts, and interactive workshops. The sets of slides, handouts, and required readings will be available from the course website. Readings are set for each week. Everyone must come to class well prepared, and ready to discuss the week’s readings with their own critical reviews and questions.

Assessment

There are three assignments for this class, which must be submitted through the course website (with no option for submission via email or physical delivery):

Assignment 1: Critical Analysis of Online Platforms or Tools —due Week 3 (20%)

Compose a 500-word analysis regarding how your chosen online platform or tool is shaped by various stakeholders’ values, business strategies, profit motives, user interface designs, and other related actors. This assignment aims to encourage critical thinking about the nature of those platforms or tools.

  1. Identify and Introduce the Online Platform or Tool: Choose a specific online platform or tool within the digital landscape and provide a brief introduction to its purpose and functionality.

  2. Analyze Influencing Factors: Investigate and analyze how various stakeholders’ values, business strategies, profit motives, user-interface designs, and other relevant actors contribute to shaping the information available on your chosen online platform or tool. Consider the impact of these factors on biases in the information available on those platforms or tools, as well as on the development, features, and overall functionality of the platform.

Assignment 2: Critical Reflection of Online Platforms or Tools —due Week 6 (20%)

Compose a 500-word reflection on the implications of the shaping factors identified in your first assignment. How do these elements influence user experiences, content presentation, and users’ overall interactions with the platform? Then, propose questions that could lead to further inquiry and a deeper understanding of the platform's or tool's nature and affordance. Consider questions related to ethical considerations, user empowerment, and potential improvements.

Assignment 3: Submit a 1000-word final paper and present your work using your chosen data analysis and visualization tools —due Week 10 (50%)

Full details of these assignments will be made available in class, and on the course website. The final 10% of your final grade will be awarded based on the quality of your class participation.

Adaptability

The course can be designed as customizable modules depending on specific goals and agendas, such as online hate speech, fake news, filter bubbles, and commercial content. The course can also be designed differently depending on the age and learning environments of the learners.

It can be re-designed for K-12 students by creating a condensed version using simplified language and content suitable for middle and high school students, taking into account their developmental stage and educational background. This adaptation may involve using clearer explanations, avoiding advanced terminology, and focusing on fundamental concepts relevant to younger learners, as the current lesson plan is tailored for individuals above the undergraduate level. It can also include more interactive quizzes, games, and real-world applications. To modify a course intended for K-12 students, the learning objectives and outcomes need to be simplified to align with their developmental stage and educational needs. The course therefore will focus on fundamental concepts and practical uses, ensuring that the topics are relevant to younger learners. Classroom activities can include interactive games, simulations, and hands-on experiments, as well as multimedia resources such as educational videos and interactive presentations.

Additionally, it can be designed as a self-paced program. It can also be designed as a program to prepare individuals to deliver the course to their specific audiences. Lastly, the course can be designed differently in various countries, incorporating examples and case studies from different regions to address local nuances in information bias.

If this course is adapted beyond the United States, for instance, in a country such as South Korea, it is crucial to match the educational objectives and results with the local educational standards and curriculum requirements unique to the Korean setting. It is important to incorporate cultural elements by including activities that are relevant to the interests and background of Korean students, using examples and scenarios that mirror their daily life and experiences. Thus, it can pose questions and initiate discussions that reflect Korean cultural viewpoints and societal norms, encouraging students to share their opinions within their cultural framework. Collaborating with local educators and stakeholders can further enhance the adaptation process by offering valuable insights and perspectives.

Reflection

A brief look back at the successes and challenges that took place during the lesson(s) design, implementation, or assessment process. What might you do differently in a future iteration of the lesson? Please also share any guidance that could be helpful for others who implement your lesson.

Acknowledgments

This syllabus and lesson plan are inspired by and refer to elements of Professor Mortiz Hardt's CS 294: Fairness in Machine Learning and Professor Solon Barocas's INFO 4270: Ethics and Policy in Data Science.

Comments
0
comment
No comments here
Why not start the discussion?