Rights : Human Rights

Value Power , Knowledge

Data - Some of the data the algorithm is based on might come from what people search on the internet - in this case, for example, celebrities who are considered beautiful AND have lighter skin, and search terms people use and suggested when searching about beauty-related information, products, and faces on advertisements.

Values - This would have a negative impact in terms of affiliation & belonging for those who are made to feel like they don't fit in the suggested "beauty standards".

Rights - When you are seeking a job that requires facing customers or public, or in industries where "good" appearances are considered high priority, you might feel less confident about getting the job because you feel like you do not fit into a category that is considered "beautiful".

Possible solution - Admit that we are all biased and implement a review of data that is used for such an algorithm. But I have to say that I personally cannot support beauty contest as beauty is subjective. Such contests are one of the reasons why people develop unnecessary biases.

The deficiency of the algorithm here comes from the lack of diverse inputs into judging the beauty of the contestants. The winners, according to the AI system, were mostly white, with Asians and Black women being left out.

A clear solution for me is widening the diversity of the people being asked, encouraging a larger variety of responses in terms of beauty.

1. Urge companies to either open-source their algorithms or make them accessible for independent auditing. This would enable third-party experts to scrutinize the decision-making process and identify any biases.

2. Provide training and education for judges, lawyers, and law enforcement officers on the limitations and potential biases of algorithmic tools. This training should equip them with the knowledge to understand when and how to appropriately use these tools.

Automated moderation leads to platform users finding alternative phrases or symbols to represent the content that they would share anyway.

Automated moderation could affect the right to health and safety, particularly at work, by suppressing content created by workers or unions intending to highlight employment rights violations. ‘Shadow banning’ on social media reduces the reach of content that automated moderation deems unacceptable. This is particularly problematic as it is hard to tell what mechanisms are in place, and the underhand, secretive nature of what content or accounts are being shadow banned means that raising an objection with the platform is almost impossible.

https://www.washingtonpost.com/technology/2022/12/27/shadowban/

The value of hedonism may be impacted by automated moderation by censoring humorous content intended to parody or protest. 

If criminal record data was an input data source for automated moderation on platforms, there would be a risk that a binary ‘has criminal record’ flag would be used to make decisions, when there is a broad scale of severity of criminal offences. People who have a criminal record as a result of peaceful protest may see their content censored as heavily as someone convicted of sexual assault. 

Some potential solutions to address these issues would be to:

• Make sure that there is always a minimum level of human moderation to check automated moderation (looking at you Twitter/X)

• Make sure that users are always alerted when their content has been removed, or they are being shadow-banned, and are the option to request that a human checks the decision 

• Train algorithms on a broad range of content that includes parody content and educational content

Actions to resolve this burning issue would be to deploy more case handlers to monitor hate speech and to effectively monitor the underlying processes. In particular every social media company should be self-committing itself to having checked a reported issues within 24h. If this is not working a respective law needs to be introduced.

Data: Personal Information

The data used to train the Beauty.AI algorithm likely included photos of participants, which might have also been paired with personal details such as ethnic identity, social media identities, and places of education and work.

The algorithm was trained on crowd-sourced images, which likely reflect societal biases. The personal information, including ethnic identity and appearance, could have influenced the data in ways that perpetuated discriminatory beauty standards, despite the algorithm's intent to be neutral. The use of data that reflects real-world biases can inadvertently embed those biases into the system.

Values: Power

The use of an algorithm to judge beauty brings with it a significant amount of power—power over how people are perceived and valued, and power over the social and cultural standards that influence people's self-worth and identity.

The algorithm's outcomes reinforced the power of Western beauty ideals and failed to account for the diverse perspectives and cultural variations in standards of beauty. The application of power through technology in this way can marginalize certain groups (in this case, those with darker skin) and undermine efforts toward diversity and inclusion.

Rights: Human Rights

There are several fundamental human rights at stake in this context, most notably the right to non-discrimination and the right to respect for private life.

The algorithm’s outcomes discriminated against people with darker skin, denying them the opportunity to be recognized as beautiful according to the system's criteria, thus violating the right to non-discrimination.

If participants' personal data (including ethnicity and identity) was used in ways they didn't fully consent to or understand, it could violate their right to privacy. Furthermore, the algorithm’s use could undermine individuals’ sense of identity by imposing a narrow, biased standard of beauty.

Potential Solutions:

1. Bias Audits and Data Diversification
   

2. Transparency and Consent
   

3. Algorithmic Accountability
   

4. Inclusive Standards of Beauty
   

5. Promotion of Diversity
   

Values - security: Many people value their security and privacy highly, but it is something that more and more people feel is being compromised in an ever-online age. More and more data-driven advertisements are being shown and this can feel like it compromises ones privacy. Again, my compromise to this would be an opt-out policy. There should be more accessible information about the how data is being used to form the personalisation of our feeds and ads, and a greater emphasis on the fact that it is a choice rather than something which is non-negotiable.

Rights - data protection: Many people use online spaces as a private place and feel that their data is not being used. However, this is not the case. Again, there needs to be more awareness made of this fact and an increased ability to not have certain data stored.

The core ethical problem is that algorithmic systems, through biased data, perpetuate and potentially amplify existing societal prejudices. This not only reflects but also reinforces discriminatory attitudes, leading to adverse impacts on marginalized groups.

Potential Solution

1. Bias Detection and Mitigation:

   • Implement regular audits of algorithms to detect and mitigate bias.

   • Use diverse and representative datasets to train algorithms.

2. Transparency and Accountability:

   • Ensure transparency in how algorithms are developed and used.

   • Establish accountability mechanisms for discriminatory outcomes produced by algorithms.

3. Inclusion of Ethical Principles:

   • Integrate ethical principles in the design and deployment of algorithms, prioritizing fairness and non-discrimination.

4. Regulation and Oversight:

   • Develop and enforce regulations that mandate non-discriminatory practices in algorithmic systems.

   • Create oversight bodies to monitor and address algorithmic biases.

By addressing these areas, we can work towards more equitable and fair algorithmic systems that align with the values of scientific objectivity, evidence-based decision-making, and the protection of equality rights.

In this sense, the use of algorithms can be an important support for those in power, so that in the face of fake news the responsibility lies with the code, not directly with anyone in particular. Although data management policies, such as the one presented on FB, required a review of the data policy and its legal management, there is a lack of progress regarding the use of algorithms in the area of ​​news management, especially when with the use of AI, it is very easy to present false information with formats and wordings that make it credible.

Algorithmic decision-making in the justice system was introduced with the promise of making sentencing, bail, and parole decisions more efficient and objective. However, reality has shown a much more troubling picture. A 2016 ProPublica investigation found that COMPAS, a risk assessment tool used in the U.S., systematically overestimated the risk of Black and Latino offenders reoffending, while underestimating the risk for white offenders. These racial disparities raise serious ethical concerns, particularly when decisions about a person’s freedom and future are at stake.

At the core of this issue is the data being used. Risk assessment tools rely heavily on criminal record data, which includes prior convictions, parole status, and time served. However, historical policing practices have disproportionately targeted racialized communities, leading to overpolicing and higher recorded reoffending rates. When an algorithm learns from biased data, it perpetuates systemic injustices, reinforcing the very disparities it was supposed to eliminate.

Another key factor is security and stability, which are often cited as justifications for these systems. The goal is to predict and prevent crime, but when the method itself is flawed, it undermines public trust in the justice system. If algorithms unfairly label entire communities as “high risk” it does not create stability- it deepens social inequalities. At the same time, this directly clashes with equality rights, such as those protected under the Equality Act 2010, which prohibits discrimination based on race, age, gender, and other protected characteristics.

How Can We Fix This?

1. Greater Transparency & Accountability - The criteria these algorithms use should be made public, and individuals must have the right to challenge their classification if they believe they have been unfairly labeled.

2. Bias Reduction in Training Data - Instead of relying solely on historical criminal records, these models should incorporate broader social and behavioral factors, such as access to education, employment history, and rehabilitation efforts.

3. Independent Regulation & Oversight - Governments should establish independent ethics boards to regularly audit risk assessment algorithms, ensuring compliance with anti-discrimination laws.

4. Human Oversight in Decision-Making Judges and parole officers must treat these scores as advisory tools, not final decisions. Context and human judgment must always be factored into sentencing.

Without meaningful reforms, these systems risk becoming a high-tech reinforcement of racial bias, rather than a tool for justice. The challenge is to find a balance- can we use AI to improve efficiency without compromising fairness?

The online international beauty contest judged by the Beauty.AI algorithm exhibited significant bias, favoring light-skinned individuals over those with darker skin. This bias arose because the algorithm was trained on a dataset that overrepresented lighter-skinned individuals, leading it to equate beauty with lighter skin tones. As a result, 43 of the 44 contest winners had light skin, reinforcing harmful societal biases that associate prestige, respect, and visibility with lighter skin while marginalizing darker-skinned individuals.

This issue violates the right to non-discrimination by embedding racial bias into an automated system, perpetuating social inequities.

Solution:

To mitigate such biases, it is crucial to ensure diverse and representative training datasets when developing AI models. Implementing fairness-aware algorithms, conducting bias audits, and integrating ethical oversight in AI development can help prevent discriminatory outcomes. Regulations and transparency in AI decision-making processes are essential to uphold fairness and inclusivity in automated systems.

Ethical Analysis:

The issue arises when Facebook’s algorithm, by automating news selection, prioritizes unverified information that can influence users’ political and religious beliefs. This impacts their right to accurate and unbiased information and may lead to discrimination or polarization. It also affects the value of personal expression, as users form opinions based on potentially false information.

Potential Solution:

• Algorithm Transparency: Facebook should explain how news is selected so users understand potential biases.

• Fact-Checking: Implement fact-checking processes before promoting trending news.

• Source Diversification: Ensure news is sourced from multiple outlets to avoid political or religious bias.

• User Control: Give users more control over the topics they want to see, aligning content with their values without algorithmic manipulation.

Data card: criminal record

Value card: affiliation and belonging

Rights card: Equality Rights, prevention of discrimination by law

Consider the ethical problem in light of the data, values, and rights information.

Suggest a potential solution to the problem.

Given access to reams of personal and contextual information about their customers via social media, companies will always be tempted to use this data to personalise prices, quotes and policy offers. To some extent, we are comfortable with data informing price: safe drivers who have no crashes or penalties on their driver’s license already receive cheaper car insurance and we accept this as a society. But more targeted and personalised alternatives still make us uncomfortable as they feel like an invasion of privacy (despite this information being publicly available and posted by us by choice).

In light of the data card being criminal record, this feels like potentially relevant information: driving offences, for example, would be relevant information to the insurance company.

Affiliation and belonging help to explain some of our reluctance to be ‘scored’ for insurance offers: we want to feel accepted as who we are by others, not judged and appraised.

And the Equality Rights Act explains well why this judgement feels discriminatory; because the insurance company is presupposing that if we have crashed our car in the past, we’ll do so again - we are judged to be a careless or dangerous driver, a group we do not care to feel an affinity with, and we are given no benefit of the doubt that we will improve as drivers and ‘deserve’ a better rate because we have changed since our driving offence occurred.

A potential solution? More social dialogue about how personalised we want our services and offers to be - and consistency on how we expect companies to apply our personal data to the products they offer to us. Because as long as there is an incentive for the company to do this (making more money from the customer, attracting more customers via more attractive rates) they will do so.

DATA
LOCATION

• mobile phone GPS

• WiFi networks

• Internet Protocol (IP) addresses

• Billing locations

• RFID (radio-frequency identification) tags

• contactless payments

• smart travel cards

• fitness trackers

VALUES
SCIENCE & KNOWLEDGE
The following values form a science and knowledge framework:

• evidence/rigour

• challenge and complexity

• different forms of knowledge

• problem-solving

• acquisition of knowledge, skills & practices

• discovery of new knowledge

• learning

• innovation

RIGHTS
EQUALITY RIGHTS
The Human Rights Act 1998 protects people in the UK from discrimination.
The Equality Act 2010 provides a legal framework to protect the rights of individuals and advance equality of opportunity for all.
You cannot be discriminated against because of your:

• age

• disability

• gender reassignment

• marriage and civil partnership

• pregnancy and maternity

• race

• religion or belief

• sex

• sexual orientation

Location data has a huge influence on algorithmic justice - whether you think that data is being collected and analysed (mobile phone GPS) or whether it's something less intrusive that you might not think of traditionally harvesting your info (fitness tracker). This location data may result in the overpolicing of certain areas, especially if those people who are tracked are those that have already been arrested/convincted of a crime, exacerbating already inbuilt bias.

The valuers of science and knowledge almost created these problems in the first place - the justice system was alreayd felt to be racist, therefore why not hand over the allocation of policing resources to an algorithm that can hopefully remove bias and treat people in a uniform way? Unfortunately this is useless if the data on which the system is trained is already racist.

Equality rights play into this in an obvious way - people deserve to be treated the same independently of their racial identity.

Solutions -

• Fall back to the already imperfect human policing systems that existed in the past.

• Remove protected characteristics from these systems - this however risks the system being able to guess protected characteristics by proxy when analysing other data.

• More human oversight and intervention in systems - use the systems bvut don't follow them blindly. Have appropriate intervention steps in the system's suggestions where it can be overridden and bias can be identified. regular independent evaluation of the algorithm's performance by a third party to ensure biases aren't being perpetuated or getting worse.

This experience was further highlighted by a recent Facebook post I made, where I delved into the subject of prejudice based on a real-life scenario. The reactions I received from my audience were telling and prompted me to reflect on the broader implications of such attitudes in our society.

Observing the responses to my post not only reinforced the importance of discussing these ethical issues but also underscored the need for greater awareness and sensitivity in our online interactions. Engaging with these topics can help foster a more respectful and inclusive digital environment.

The post received a range of reactions — some thoughtful, others deeply negative, even humiliating. 

What struck me most wasn't just the content of the responses, but how they reflected a pattern: 

➡️ The most negative comment received 61 likes — more than any other thoughtful or constructive reply. 

That’s not just social behavior — it’s part of a feedback loop. 
Platforms like Facebook track what gets engagement, and that data becomes part of what the algorithm promotes. I chose the Data Preference card to coincide with the above

Unlike people, AI has no empathy. It doesn’t understand shame, context, or compassion — it just learns from the data we feed it. 

When considering the Value card, I prioritised Security, particularly in the context of predictability algorithms that often depend on historical data. These algorithms can perpetuate biased outcomes, as illustrated by experiences where preferences are manipulated to create feedback loops. This tracking of engagement can lead to a scenario where negative posts are amplified, resulting in an increase in adverse comments and ultimately affecting individuals' (My) mental well-being. 

That’s why I believe we need to talk about bias more openly — not just in how humans treat each other, but how our reactions are manipulated, recorded, and turned into the next generation of "intelligence." 

Now, was the bot unfair? Or course it wasn’t the bot’s fault, the developers didn’t anticipate (or ignored) how toxic online spaces can get.  In this case, the lack of caution led to ethical misconduct.

On the other hand, looking at the values from the image: fun, humor, entertainment, they were completely misused here. There’s a line between humor and harm, and these users jumped over it without hesitation in the case. When people’s fun involves corrupting a system to spew hate, you’re not just joking, you’re contributing to a toxic digital culture, contributing to weaken the values.

Ethically, both sides failed, the chatbot’s designers didn’t protect their AI tool or think deeply enough about the consequences, and users ignored basic decency. Now, can we say that technology reflects us? in a certain point yes, but I think that unlike humans, artificial intelligence would not and should not be hypocritical.

After having studied the case I was wondering about what does an online AI chat understand in this case: and its answer was:
How did I interpret the fairness of the case?

·        Users intentionally fed Tay biased, harmful, and unethical content to corrupt its learning.

• This is an example of irresponsible use of digital platforms, violating integrity, respect, and digital citizenship ethics.

Thus, the AI chat said that the chatbot failed ethically by design, the user behaviour was also unethical by intent. What do you think about this story?

At the heart of this issue are some really important values and rights. Every student deserves to feel safe, respected, and included at schoo,no matter who they are. Hate speech and online bullying go against these basic rights and the values of fairness and kindness that schools should stand for. That’s why it’s so important for schools to have clear rules about how students treat each other online, to teach about respect and empathy, and to make it easy for students to speak up if something’s wrong. When schools take these steps, they’re not just following rules—they’re showing that everyone’s dignity and wellbeing truly matter.

The fake news epidemic is deeply intertwined with personal information, power imbalances, and human rights. Some key conversations through this lens are:

1. Personal Information: There is a vast amount of user data (browsing habits, interests, demographics) harvested by platforms and malicious actors. This enables micro-targeting, or the fabrication of misleading content tailored to exploit individual biases, fears, and identities, making it highly persuasive and difficult to resist.

2. Power: There is also the inevitable presence of power; one example is how data asymmetry creates immense imbalances through different stakeholders like platforms & advertisers, who monetise attention, often prioritising engagement (even from falsehoods) over truth, fuelled by personal data. In contrast, individuals lack control over their data and how it's used to manipulate them, feeling powerless against the flood of disinformation.

3. Human Rights: Ultimately fake news directly threatens fundamental rights:

   Like the Right to Information (Art. 19, UDHR): fake news pollutes the information ecosystem, hindering access to accurate information necessary for informed decisions.

   Freedom of Expression (Art. 19, UDHR): Can be used to drown out legitimate voices, incite hatred against minorities, and chill speech through intimidation. Another article worth referencing in the UDHR convention is the rights to life, security, and health (Arts. 3, 25, UDHR): the spewing of dangerous health misinformation during the pandemic was a big worry that led to incitements and fatal consequences for many people.

Talking solutions: a kind of Layered accountability & empowerment

combating fake news might be worth considering, because curbing fake news requires a multi-pronged approach addressing all three lenses:

1. Strong Data Protection & Privacy Laws: Limit the collection and exploitative use of personal data for micro-targeting (e.g., GDPR-style regulations globally). Empower users with genuine control over their data. Also a transparent & accountable platform design that could combine algorithm transparency as a requirement for platforms and a design feature that slows virality: by propting before sharing unverified contents and consistently display creibility ratings might be solutions worth discussing.

My data card is around collecting data on personal habits. I think this could be parsed two ways in our scenario - the data of the bullied and the bullies. It cannot be considered, though, without understanding the values and rights we're trying to maintain. I have affiliation and belonging and human rights.

From a utilitarian perspective, hate speech and cyberbullying should be outlawed to preserver our ability to find membership in the community, enjoy social interaction, and fulfill our basic right to be treated humanely. Any data collection on personal habits that are used towards the end of ending cyberbullying and hate speech should be allowed.

But oh, not so fast.

Certainly, we can think of an anecdote where this all makes perfect sense. A social media company monitors and stores information about a person who is committing hate speech. That information can be shared with law enforcement to assist in his arrest, therefore making the community happier and safer.

But what if this data was collected and the contextual integrity was not respected. What if information about where we are and what we like to buy was sold and used in different context? With the contextual integrity broken, how happy would the average user be now? And what responsibility does the owner of the data - in this case, the social media platform - bare for how this information is later used? What if this information was shared with law enforcement to make low-level drug arrests? What if it was used by immigration enforcement to target particularly ethnicities?

Moreover, what if one form of hate speech was targeted more than another? What if racist speech was tolerated by law enforcement if it was against Muslims, but so-called "anti-white" hate speech was forcefully dealt with using this data?

The idea of algorithms that help us maintain our rights and values is like a wonderful dream, but the reality is a nightmarish mess of nuance. It is not, however, insurmountable.

An elected human panel of monitors who ensure the contextual integrity of data would go a long way in helping solve this problem. These panels could be public officials or people appointed to private companies as "watchdogs." Ideally both panels would exist. While any private enterprise would likely fight this action or appoint pro-business members to the panel, regulations could stipulate its required formation.

Regulatory oversight or forcing companies to take responsibility for what happens with the data they collect and sell would also help solve this issue. Instead, we absolve them of all or most responsibility, or penalize them with fines that amount to rounding errors on their balance sheets.

Problem: a firm suffers reputational damage in the wake of misuse (say, due to a lack of proper data management and governance) of its customers' behavioral data that they collect for analytical purposes.

By virtue of the set of values chosen, the best way forward for the firm would be to demonstrate "honesty and sincerity," communicating its remorse to its stakeholders and being accountable to any possible legal challenges that might go its way. The damage has been done, and if the board and further flak. Integrity also behooves the firm's board and management to take stock of the firm's values and ensure that their strategy hews to them. Going forward, they could invest more in enhancing their data strategy and pouring more resources into data governance and cyber security.

So, the ethical issue is about gender bias in Google’s ad targeting system, where women were shown fewer high-paying job ads than men. The study that showed this created identical fake profiles, differing only by gender, and still the results were unfair.

Now, health records don’t seem directly related at first glance, but if personal health data ever becomes part of targeting (say, mental health or disabilities), it could make things worse. It’s easy to imagine that certain people could be doubly excluded - by gender and by assumptions made from their health data.

Then there’s the value of hedonism - all about happiness, pleasure, fun, and personal growth. But how are people supposed to pursue those things if algorithms quietly limit their opportunities based on who they are? That goes completely against the idea of a fair and open digital environment where everyone can thrive and chase a better life.

And finally, consumer rights. That made me think: if you can return a faulty toaster, why can’t you complain when an algorithm treats you unfairly? Maybe systems like ad targeting should come with a kind of transparency label - showing what criteria they’re using, and letting users challenge or correct it when it’s clearly biased.

So yeah, strange combo of topics - but it all comes down to designing tech that actually supports people and doesn’t quietly push them out of opportunities they deserve.

From the perspective of altruism (values card), the following challenges can be identified:

• Justice, morality, and fairness: How do we determine what is good (harmless) and what is bad (harmful)? Is that judgment valid, and on what grounds?

• Diversity: Does detection and removal undermine diversity, that is, the exchange of diverse opinions?
  

From the perspective of l human rights (rights card), the following challenges arise:

• The right to personal liberty and freedom of expression: Is it legitimate to deprive users of their freedom of expression when their posts are detected and removed? What is the justification? This ties back to the first point: do we say “it is hate speech that harms others, therefore deletion is justified” or “it falls under individual freedom of expression and therefore should not be deleted”? Where do we draw the boundary between the two?
  

Finally, from the perspective of official identity data (data card), the following issue can be raised:

• Posts flagged for deletion may lead to the identification of the poster’s identity and possible intervention by authorities.

Furthermore, given the characteristics of our mobile devices, which track our activity not only through fitness apps but also via Bluetooth beacons, payment information, or simply GPS, and thanks to the consent granted on our devices, applications can monitor our movements. This data allows them to influence us by offering products, services, or content that impacts our self-perception of beauty, and specifically, our progress toward achieving it.

Many companies offer app, based services, often at no cost. They present these services with useful features wrapped in a veneer of Honesty and Credibility, but the reality is that when something is free, we are the product, our data is. Companies use this data to generate targeted communications and ad-hoc products/services that prey on the perceived inadequacy of users who don't meet the idealized beauty standard. This exploitation ultimately infringes upon people's Rights to Health and Safety, specifically their right to a healthy mental and physical life without the malicious interferences that can put their lives at risk.

A relevant real-world example is when some retail companies require job candidates to submit a video introducing themselves and showcasing their knowledge, but also their physical appearance. Certain brands ensure that all their store assistants strictly adhere to the established societal beauty pattern. This situation unequivocally highlights the need for companies to develop internal programs for 'Healthy Beauty' and communicate clear guidelines to their employees, expressing paths to achieve a healthily beautiful life and avoid these problems.