Class 4 - Addressing inequity in AI by Ekin https://padlet.com/ekiny/ycfliwynk159rdq5 Share group members names + share your solution ideas en-us 2023-04-20 16:46:08 UTC 2026-02-12 09:51:27 UTC hello@padlet.com Ishita, Ali, Ashley, Jenny https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562711297 Issue: digital divide for rural and low-income communities

Actions: 
1) Someone in power needs to care
2) resources and partnerships with tech and fiber optic access
3) contribute to data collection
4) build access in all languages]]> 2023-04-21 03:08:05 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562711297 Team Members: Baruni, Samiksha and Roshni https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562711749  Issue: Accessibility to Education
-> Pushing for awareness of digital tools for all despite societal status
-> Make technological devices affordable for all 
-> Upskilling the older generations and raising awareness amongst the youth on how to ethically use AI
-> Regulate the use of AI

Equity for all:
-> Actively work on removing biases on  datasets on which AI is trained
-> Ensures inclusive hiring practices across organizations 

 

]]> 2023-04-21 03:08:33 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562711749 Teams: Coldie Wu, Shiyang Li, Guangdi Zhu, Handan Xiao https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712107 The first bias thing jumped in our mind is that when using an open AI which can help you to translate text to image, when you entering "nurse", it always comes out with female nurse instead of male. 
We think one of the biggest reason why AI has the bias is that the designing and the data used for training AI are not equity.
Here are the solutions we come out:
- trying to make sure that the researcher and the developer of the AI do not have such bias, they can filter the original data that used for training AI does not have such bias.
- making sure the date resource is gathered from different communities, full of diversity.
- correcting AI every time when it comes with bias.]]> 2023-04-21 03:08:53 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712107 Yuze Li, Yuqi Shi, Yuxin Li https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712130 What are some arenas of inequity in our society that AI could potentially broaden?

1.     Employment: Artificial intelligence could lead to the automation of many jobs. Therefore, this will allow unskilled workers to be replaced by AI, and those workers will lose their jobs and have no income. At the same time, this will lead to higher unemployment and lower transaction rates in the society, as many families lose their source of income.
2.     Education: Artificial intelligence will cause some students to use artificial intelligence instead of thinking by themselves. At the same time, artificial intelligence will also deepen the inequality of education. Some students cannot use artificial intelligence, which will be unfair to these students. In the same way, it is unfair for students who use AI to learn nothing and to be unable to think, despite their own choices. In general, artificial intelligence is of no benefit to education.

Think of examples where AI could amplify existing biases or discrimination, or where it could unintentionally create new forms of inequity.  
1.      Lack of variety: The algorithms of AI systems can lead to a lack of results. This can therefore lead to a lack of understanding of the different needs of different communities, leading to the development of AI systems that do not serve everyone equally.
2.     Stereotypes: AI systems may inadvertently reinforce stereotypes. For example, a recruitment tool that uses artificial intelligence to filter job applications might give preference to male candidates over female ones. Because its algorithms are trained on historical hiring data that show a bias against men in certain jobs.

1-2 ideas for an actionable solution idea based on Communication and Leadership.  How can we make sure that AI tools do not deepen existing inequities, leaving some communities even further behind?  What would be a solution?

1.     Promote diversity in artificial intelligence.
2.     Evaluate and upgrade AI periodically, to avoid the occurrence of Stereotypes.
3.     In the case of education, the vetting system for screening AI in education should be strengthened. For example, every paper handed in is reviewed to see if there is any trace of AI. This can avoid the problem of students not thinking independently.
4.     In terms of employment, research on AI can also be reduced to give more people job opportunities. The money spent on AI could be used to pay workers. Ensure the equal development of society.

]]> 2023-04-21 03:08:54 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712130 Isaac, Manon, Rachael, Vivian https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712231 Issue: Transparency. 

Actions:
1. Educating people how to use ChatGPT more effectively. 
2. Put more diverse Data into ChatGPT.]]> 2023-04-21 03:09:00 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562712231 Haitao Ji, April(Xiangyuan) Jiao, Yubo Fan https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713129 The development of AI may lead to inequity in the job market. For some low-skilled workers, their jobs may be replaced by AI, reducing their job opportunities and work experience, which would make it harder for them to find the next job, causing greater unfair treatment.

Possible solutions to address this issue could include:
  1. Holding job training events to promote and educate workers on how to better use AI to improve their work efficiency, allowing AI to help them do their jobs better.
  2. Organizing community forums to allow low-skilled workers to exchange and learn vocational skills and legal knowledge, improving their skills while safeguarding their legal rights to ensure that they receive fair compensation in the event of layoffs.
]]> 2023-04-21 03:09:52 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713129 Group: Kirstie, Mary, Dylan https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713846 What are possible inequities that could be caused or broadened by AI?
-People without access to technology and ai like chat gpt are at a disadvantage in school and the job field, thus exacerbating existing inequities 
-Companies using AI generated images to make themselves look more diverse or inclusive 
-Biases within data sets, creating bias in AI response

What are actionable solutions?
-Diverse perspectives in workplaces, and creating AI 
-Increase accessibility and education to AI 
-More diverse perspectives and data sets
-Consider social implications of AI use before deploying new programs 
-Diverse leadership within AI teams
-Expand access to all locations and languages

]]> 2023-04-21 03:10:31 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713846 Melissa, Bella, Keri https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713876 Issue: Bias in AI systems with resumes 

Solutions:
  • Implementing a worker's bill of rights 
  • Updating AI system at company on a regular bases and running it through test groups 
  • Being mindful of locations and creating AI that provides equal access to information regardless of location
  • Transparent communication and detailing the AI each company uses
  • Having forums for the public to input their experiences and then making changes
]]> 2023-04-21 03:10:33 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713876 Ian, Yuzhi Lu, Cola Wang, Leopoldo https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713946 Criminal Justice: AI-powered risk assessment tools used in the criminal justice system may inadvertently perpetuate racial and socioeconomic biases. For example, if the algorithm is trained on historical data that reflects biased policing practices, it may predict that certain individuals are more likely to commit crimes based on their race or socioeconomic status.

->have the warning sign for minorities 
->update data frequently 
->Make proper references to avoid copyright issues


]]> 2023-04-21 03:10:37 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713946 Group: Jessie Yu, Nina Li, Joanna Yang https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713992 AI has the potential to exacerbate biases related to gender.

Recommendations:

1. It is important to continually evolve our own thinking and actively work towards eliminating gender bias. By changing our own perspectives and attitudes, we can ensure that biased beliefs and attitudes do not become embedded in AI systems.

2. When training AI, it is crucial to be avoiding the use of biased words or algorithms. We should be careful to the use of data and inputs during the training , in order to minimize the incorporation of any existing biases. ]]> 2023-04-21 03:10:40 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562713992 Alice, Matt, Jinny alicepengu https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562714329 Issue: Accessibility, Data Bias
1. Education: To increase accessibility to AI, it is important to provide education and training on how to use AI tools and technologies. This includes providing education for people who may not have a technical background, as well as for those who do. This will help to ensure that everyone has the skills necessary to take advantage of the benefits of AI.

2. Address Bias in Data: acknowledge and address the bias in data that can perpetuate inequalities. 
Encourage diversity and inclusion in data collection efforts, and use methods that minimize bias in data collection, such as random sampling. Ensure that data is collected ethically and with consideration for privacy and consent.
Regulation for Data Science: Regulations and policies to ensure that data science is used in a responsible and ethical manner. Setting healthy boundaries for the use of data science, such as requiring a certain percentage of minorities to be included in data sets. 
Global collaboration for One Data Set: Collaboration between different stakeholders, including businesses, government, and civil society, and help to create a single, standardized data set that is inclusive, diverse, and representative of all populations. (UN for AI)]]> 2023-04-21 03:10:58 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562714329 Sandhra, Sunny, Yue Zeng https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562714549 Regular audits and data set updates :Frequent data set updates and audits, such as testing the system on diverse data sets, analysing outcomes, and investigating any potential biases discovered, would help to ensure that AI tools are designed and trained to be fair and equitable, reducing the risk of deepening existing inequities and leaving some communities even further behind.]]> 2023-04-21 03:11:11 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562714549 Nick, Gabriela, Katie kapl1914 https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715038 Issue: AI tools are being used to filter job applications and resumes. These can perpetuate bias against people of color and women (if the algorithms are trained on biased data/reflect historical discrimination.)

Solution: Create a tool where AI can filter through resumes and applications while removing gender/race entities. Being able to remain a useful tool/time saver while only focusing on skills and work experience + removing any personal descriptors/names/etc. ]]> 2023-04-21 03:11:36 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715038 Dorothy, Jay, Thit Thit https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715174 The arena of inequity AI could potentially broaden:
AI-based voice assistants like Siri or Alexa have been found to faced difficulties in comprehending accents and dialects other than the typical American English. As a result, individuals who speak English as a second language or those with accents that the algorithm is not familiar with could face unfair treatment. This could lead to inequities in their interactions with voice assistants. 

Solution 1. 
Inclusive testing - before release of the voice assistant, it can be tested with a diverse group of users

Solution 2.
Transparency - make the information about how the product will crawl or use the data transparent to the public

Solution 3.
Set up a checking mechanism - set the principles for the diversity assessment

Solution 4.
Multiple language support - design the voice assistant to support multiple languages and dialects

]]> 2023-04-21 03:11:44 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715174 Danica Wood, Ping-Chun Chen, Amber Pan https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715356 Issue: Unequal Hiring & Recruitment practices for companies. Ex: Use AI to narrow down search 

Solution: Ensure the data range is used as a reference to gain more information about abilities / skills but not causing vital issues that can mislead recruiters. Some companies use AI now and it eliminates potential candidates because of certain wording. Therefore, broadening words / skills that can be translated for certain jobs. Example, maybe having companies provide examples and allow AI to find skills / positions that may be applicable. 

It is also important for companies to have clear transparency and explanations for recruiters and applicants to understand the technology and what the AI process is for candidates. This provides a fair opportunity for all applicants to understand why they may or may not be chosen in advance. In addition, help individuals understand how AI works, and a lot of individuals do not want to use it / trust it for that reason. Example: Companies using AI right now may eliminate individuals who do not have specific words on their resume, so being transparent about what the AI is searching for in the initial process can help individuals make edits / changes to their resume as needed or know that they are not qualified for a position.


]]> 2023-04-21 03:11:56 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562715356 Isobel, Josh, Jason https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716159 One area of inequity that AI could potentially broaden is the idea of job displacement, especially in lower-income fields and the service industry. Potential solutions could be: 
- A business framework that emphasizes and values human connection and outlines the need for interaction in jobs
- Organizations re-skilling their employees in a variety of areas to ensure that their commitment to a company can be visualized in multiple areas
- Government regulation to ensure organizations commit to a varied and substantial workforce and internal movement to other positions. ]]> 2023-04-21 03:12:46 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716159 Nayeli, Allison, Nadir, Shraddha https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716376 Issue: Underfunded schools in rural regions with lack of resources and educational data
Solution: Provide tutoring and other educational assistance through their provided tech. Being accessible in libraries and schools, embedded already for people who may not have access to internet. Using diverse data, the use of chatbots can provide support to students who need it.]]> 2023-04-21 03:12:59 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716376 Wenhui Li, Yuwei Cui, Lei Cai, Hanxiao Zhang https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716507 AI-based education systems can perpetuate biases if they are trained on biased data, such as recommending courses or majors traditionally taken by men, even if a female student would be more successful in a different course or major. To mitigate this issue, the system's training data should be diverse, representative, and inclusive. Regularly evaluating the system's recommendations, making the algorithms transparent, and collecting user feedback can also help identify and eliminate biases in the system.
By implementing these solutions, AI-based education systems can become more inclusive and equitable. They can provide personalized recommendations that cater to the diverse needs of all students, while eliminating biases and discrimination that perpetuate stereotypes.


]]> 2023-04-21 03:13:08 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716507 Amy Huang, Amber Cheng ,Daisy Zhu, Janie Zhong https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716680 Issues:
1.Hiring and Employment: 
AI-powered recruitment tools may unintentionally discriminate against certain groups based on their gender, race, age, or other factors. For example, if an AI model is trained on historical data that is biased against women or minorities, it may perpetuate those biases in the recruitment process. Similarly, AI-powered performance evaluation tools could create inequity if they are not designed to account for differences in cultural backgrounds, disabilities, or other factors.

2.Criminal Justice: AI-powered risk assessment tools are being used in the criminal justice system to predict the likelihood of reoffending, but these tools may unintentionally amplify existing biases. For example, if the historical data used to train the AI model is biased against certain groups, the tool may be more likely to flag members of those groups as high-risk, even if they are not. This could lead to unfair sentencing and further perpetuate inequity in the criminal justice system.

3.Healthcare: AI-powered diagnosis and treatment recommendation systems could unintentionally perpetuate health inequities if they are not designed to account for differences in access to healthcare, socioeconomic status, and other factors. For example, if an AI model is trained on data that is biased towards certain demographics, it may be less accurate in diagnosing conditions in other groups, leading to delayed or incorrect diagnoses.

4.Financial Services: AI-powered credit scoring systems may unintentionally discriminate against certain groups based on their race, gender, or other factors. For example, if an AI model is trained on data that is biased against minorities, it may be more likely to deny loans or credit to members of those groups, perpetuating existing economic inequities.

Solutions:
In our view, it is essential to ensure that AI is designed and developed in a way that does not perpetuate existing inequities or create new ones. This requires careful attention to the data used to train AI models, as well as ongoing monitoring and evaluation of their performance to ensure that they are not perpetuating biases or discriminating against certain groups. Additionally, it is crucial to involve diverse voices and perspectives in the development and deployment of AI systems to ensure that they are designed in a way that is equitable and serves the needs of all members of society.]]> 2023-04-21 03:13:18 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562716680 Sean, Malcolm, Joel https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562717904 Issue: Global inequity in access to AI tech, harm from AI biases (surveillance, ie)

Solution: Create a F.A.T.E. based framework (Fairness, Accountability, Transparency, Ethics) between countries outlining fair use and the commitment of member countries to provide development money for developing countries to expand their AI capabilities. 

Issie: Loss of employment to automation, loss of worker access to wealth generated by AI. 

Solution: Major legislation (or perhaps constitutional amendment haha one can dream) guaranteeing worker rights and/or housing and income guaranteed by the government so that we can more equitably transition to a post-work society. ]]> 2023-04-21 03:14:29 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562717904 Team: Janie, Taylor, Nibedita https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562718671 Issue: AI can widen the gap of data inaccessibility, where people in the developing countries might not have access to its usage.

Solutions:

- Giving people in the developing countries more internet access, where they participate in AI's data collection.

- Using AI to discover where there are populated areas in developing countries, where internet cafes can be built.

- Owners of the internet cafes hosting AI training sessions for users, so they can become aware and familiar with using AI as a creative/productive tool.]]> 2023-04-21 03:15:14 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562718671 David Liang, Elmo Li, Michelle Xu, Judy Deng https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562718710 Potential Issues
Data bias:
  • Loan approval algorithms may deny loans to people of color due to biased historical data.
  • Language translation algorithms may reinforce gender stereotypes based on biased data.
Amplification of existing biases:
  • Predictive policing algorithms may unfairly target people of color due to racially biased training data.
  • AI tools analyzing job applications may discriminate against candidates based on their race, gender, or age.
Lack of transparency:
  • Credit scoring algorithms that use social media data may be opaque, making it unclear why credit is denied.
  • Predictive analytics tools used by law enforcement may be opaque, making it difficult to understand how decisions are made and who is targeted.
  • AI-powered recommendation systems on social media may be non-transparent, making it unclear why certain content or products are recommended. Which then causes concerns about content censorship to certain groups of people especially teenageers.

Actionable Item
AI-powered recommendation systems on social media may be non-transparent, making it unclear why certain content or products are recommended. Which then causes concerns about content censorship.

Solution
  • Improve transparency of recommendation algorithms.
  • Establish clear guidelines for content.
  • Conduct regular audits of recommendation algorithms.
  • Empower users with customization options.
  • Collaborate with regulators and experts to develop best practices.
  • Making data sets more transparent and publicly available can help to reduce the risks of bias and errors in the models.
]]> 2023-04-21 03:15:17 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562718710 Linh, Amelia, Hans https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562719255 Some arenas of inequity:
Education:
If some kids or schools don't have access to this technology, they might miss out on the benefits it can bring.

Hiring:
AI algorithms can be biased based on the data they are trained on. It can lead to discrimination against certain groups, such as gender or age

Solutions to minimize the inequities
Bias identification: AI developers should use these methods to prevent AI from reinforcing biases. This includes finding and fixing data and algorithm biases.

Inclusive design: AI systems should be designed to be inclusive and accessible to all people, regardless of their backgrounds or abilities.
]]> 2023-04-21 03:15:53 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562719255 Srihari and Jesse https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562721874 >Inequity in terms of accessibility. AI is a tool and people will be exposed to inequity when they don't know how to use it.

>Although ChatGPT is free now there are many other AI tools that are more catered towards specific fields which is at a cost and this might not be something that every business can adopt or afford

]]> 2023-04-21 03:18:23 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562721874 Ashwini, Ardhra, Maniti, Wei-J https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562723222 Areas of inequalities - 

  1. Limited accessibility. While most of the tools are free, they might eventually become paid. (Chat GPT already has a subscription model) In that case, wouldn’t labor for the less fortunate look difficult? 
  2. Biases in healthcare
  3. Information inequity during world events. 

Actionable Ideas - 

  1. Build counter AIs that are trained to detect and snub fake news or spread misinformation, especially around events such as elections.
  2. Using AI to detect bias in the text that people write without using AI. For example - what if Grammarly was heavily trained on datasets to detect bias, and so when a person uses it to pen down their thoughts and writes something that is biased, Grammerly cleverly suggests a reframing of the text? (Might mess with freedom of speech) 
]]> 2023-04-21 03:19:44 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562723222 Wendy Yang, Flora Fu, Lucy Hu https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562726083 The particular case we discussed in our group is how AI generated algrerasim evaluate and search candidates for job and specific roles which might cause inequality. We want to know the question about how these algrorism calculate and how biased that might be based on the current standard to select candidates which might continue have an impact on future selection. 
Some of the ways we purposed: 
  1. Not have only one way of standard but see diversity among candidates so that AI learns to select candidates not just based on one standard for all.
  2. Be able to have transparency on how these algrosim are calculated and provide ways to manipulate them to make sure it provides even chance among different candidates.
]]> 2023-04-21 03:22:28 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562726083 Biying Wu, Yujia (Gia) Zhai https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562739680 AI is undoubtedly poised to bring about significant changes to society in the near future. However, along with its potential benefits, there are also social problems that we can foresee. Let's discuss a couple of them in more detail:

Gender Inequality: AI algorithms, if trained on biased data or designed with biased assumptions, can lead to discriminatory outcomes, such as biased hiring practices, gender-biased language processing, and biased decision-making in areas like lending, insurance, and criminal offenses. 

--->Solution:We need to ensure strict control over data quality, database sources, and data cleaning process. We should also prohibit AI from making absolute judgments and ensure that all policies and conclusions adhere to algorithmic fairness and inclusive design principles.

Geopolitics: The impact of AI extends beyond social awareness. In our highly globalized world, countries with varying levels of AI development and access to AI resources may face economic, social, and political disparities, potentially leading to geopolitical tensions. For instance, issues related to industrial upgrading, data sovereignty, and surveillance have already been contentious topics due to unequal national power. The differing attitudes of countries towards AI development may even impact social development in the next century. 

--->Solutions: To tackle these challenges, international organizations need to play a role in promoting international cooperation. It's crucial to cooperate through organizations and consensus frameworks for ethical AI use. Strong regulations should also be in place to ensure responsible and fair AI practices globally, preventing an unhealthy arms race-like scenario in the future.]]> 2023-04-21 03:36:58 UTC https://padlet.com/ekiny/ycfliwynk159rdq5/wish/2562739680