The Significance of "Clean" Data And Bias.

Algorithms for machine learning are only as good as the data used to train them.

Making sure the data you are using is as clean as possible is a crucial aspect of governance. 

We typically refer to data as being "clean" when it is both of high quality and bias-free. 

Data Integrity.

Quality refers to a set of measures that may be used to judge whether data is appropriate for the purpose for which you want it. 

To make sure you're meeting all the requirements for data quality, each of these factors should be examined since they are all equally significant. 

Consistency is the first one. This indicates that a dataset's data was all collected and recorded in the same manner. 

For instance, if a record has numerous fields, each record should have each field filled in. 

In order to always be able to utilize the data together, fields should be used consistently across all records and if we "know" anything about one item of information in the dataset, we should also "know" it about all other pieces of information. 

On to accuracy. This only indicates that the data is devoid of errors. 

The tools or sensors that have been used to gather or input the observations and measurements must be audited and shown to be operating properly for the observations and measurements to be considered accurate. 

We must make sure that stakeholders who are responsible for data entry are taught and aware of all the governance standards since mistakes may still happen when data is entered by humans. 

Another crucial measure is uniqueness, which simply refers to the absence of duplicate entries. 

When processing data, it's highly probable that your database may start to become inaccurate if the same piece of information is stored more than once in different entries. 

Validity is a metric for determining whether each record or piece of information in a database is appropriate for the use for which it is intended. 

For instance, are dates kept consistently and are all numbers stored consistently—as integers or rounded up or down to a certain decimal place? Timeliness evaluates how probable it is for your data to be relevant given the period it was obtained. 

Some processes, like the movement of glaciers, may be monitored and understood with just sporadic measurements. 

Others need measurements to the millionth of a second, such the locations of protons and electrons in a subatomic structure. 

Measurements must be collected and recorded with a delay that is as near to zero as is possible for procedures that call for real-time datasets. 

Last but not least, completeness measures how much of the entire amount of data that is available on a topic is represented in your dataset. 

Make sure every item you carry in your inventory is reflected in the database if you're utilizing a database of your items and pricing to determine which are the most popular. 

It would not be possible to capture the whole information for other uses, such as tracking animal migration routes, thus a sample would be selected for tracking and analysis. 

However, your insights will be more realistically based the more full your collection is. 

Data governance requires auditing your data using metrics that monitor these parameters in order to confirm that you are dealing with high-quality data. 

Biased Data.

Bias is the second component of "clean data." Data that is biased is not accurately reflective of the data topic. 

Typically, this is caused by elements innate to the method of data collection. 

For instance, your data would be intrinsically skewed if you use feedback forms to gauge customer happiness and only distribute them to clients who have given favorable evaluations. 

Bias may always come in because of the enormous and intricate datasets utilized in AI and machine learning projects. 

The fact that skewed data implies your insights won't be based on objective truth is a significant hurdle for many data endeavors. 

In fact, if the promise of AI is to be fulfilled, experts in the area of AI believe that one of the main problems society will have to solve is eliminating prejudice (or at least decreasing the harm it may bring). 

As mentioned above, biased data may be the consequence of poor data quality, but bias can also appear when your data is of high quality overall. 

This is due to the fact that bias might exist even whether the data is correct, distinctive, reliable, or current. 

It indicates that you aren't casting your net far enough to collect data from a range of sources or perspectives. 

As a consequence, the simulations and models you create won't accurately reflect reality. 

Data bias has some extremely dangerous consequences. 

When face recognition technologies used by police forces to locate criminals in crowds in the US were audited, it was shown that young, female, black citizens were considerably more likely to be incorrectly identified than those of any other age category. 

The algorithm's accuracy rate was determined to be 34% lower for this population when used in comparison to other groups. 

This may undoubtedly result in more erroneous arrests, stops, or searches of persons in this demographic if allowed uncontrolled. 

Recruitment is another instance where data bias might really be problematic. 

The datasets utilized by recruiting algorithms that Noel Sharkey has investigated are so rife with bias that they simply shouldn't be used unless they can be controlled and reviewed with the same level of rigors as is required for data used in pharmaceutical trials, according to Sharkey. 

After realizing that a machine learning algorithm used by Amazon to evaluate job applications was essentially sexist, the company ceased using it in 2018. 

The dataset that the algorithm used was found to discriminate against women, passing them over for opportunities for no reason other than the fact that it did not have enough data on female applicants for these roles because far fewer women than men had applied to work for the company over the previous 10 years. 

The fact that it may sometimes be appropriate to purposely add bias into a system in order to make up for societal elements that tend towards injustice or intolerance complicates matters even more. 

Microsoft and IBM both introduced AI-powered chatbots during the previous decade, but they eventually needed to be modified (or, in Microsoft's case, deleted) to prevent them from behaving in an offensive and bigoted way. 

This was due to the fact that they were developing their communication skills based on social media conversations, which are of course sometimes racist or hostile in nature. 

This entailed informing the bot that it shouldn't be learning from racist or abusive material, which necessitated injecting a deliberate bias aspect into the system. 

Of course, this obviously results in less realistic data being used to train the bots. 

There weren't many options, but it's obviously inappropriate for a computer speaking for a business like IBM to use racial slurs and quote Hitler. 

Another crucial step in the governance process is to balance the damage that might result from using biased data against the harm from omitting it. 

~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram

References And Further Reading

1. Roff, HM and Moyes, R (2016) Meaningful Human Control, Artificial Intelligence and Autonomous Weapons, Briefing paper prepared for the Informal Meeting of Experts on Lethal Autonomous Weapons Systems, UN Convention on Certain Conventional Weapons, April, article36.org/wp-content/uploads/2016/04/MHC-AI-and-AWS-FINAL.pdf (archived at https://perma.cc/LE7C-TCDV)
2. Wakefield, J (2018) The man who was fired by a machine, BBC, 21 June, www.bbc.co.uk/news/technology-44561838 (archived at https://perma.cc/KWD2-XPGR)
3. Kande, M and Sönmez, M (2020) Don’t fear AI. It will lead to long-term job growth, WEF, 26 October, www.weforum.org/agenda/2020/10/dont-fear-ai-it-will-lead-to-long-term-job-growth/ (archived at https://perma.cc/LY4N-NCKM)
4. The Royal Society (2019) Explainable AI: the basics, November, royalsociety.org/-/media/policy/projects/explainable-ai/AI-and-interpretability-policy-briefing.pdf (archived at https://perma.cc/XXZ9-M27U)
5. Hao, K (2019) Training a single AI model can emit as much carbon as five cars in their lifetimes, MIT Technology Review, 6 June, www.technologyreview.com/2019/06/06/239031/training-a-single-ai-model-can-emit-as-much-carbon-as-five-cars-in-their-lifetimes/ (archived at https://perma.cc/AYN9-C8X9)
6. Najibi, A (2020) Racial discrimination in face recognition technology, SITN Harvard University, 24 October, sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/ (archived at https://perma.cc/F8TC-RPHW)
7. McDonald, H (2019) AI expert calls for end to UK use of ‘racially biased’ algorithms, Guardian, 12 December, www.theguardian.com/technology/2019/dec/12/ai-end-uk-use-racially-biased-algorithms-noel-sharkey (archived at https://perma.cc/WX8L-YEK8)
8. Dastin, J (2018) Amazon scraps secret AI recruiting tool that showed bias against women, Reuters, 11 October, www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G (archived at https://perma.cc/WYS6-R7CC)
9. Johnson, J (2021) Cyber crime: number of breaches and records exposed 2005–2020, Statista, 3 March, www.statista.com/statistics/273550/data-breaches-recorded-in-the-united-states-by-number-of-breaches-and-records-exposed/ (archived at https://perma.cc/BQ95-2YW2)
10. Palmer, D (2021) These new vulnerabilities put millions of IoT devices at risk, so patch now, ZDNet, 13 April, www.zdnet.com/article/these-new-vulnerabilities-millions-of-iot-devives-at-risk-so-patch-now/ (archived at https://perma.cc/RM6B-TSL3)

Understanding Evolving AI Ethics

AI really has the power to transform business and larger society in ways that are now unthinkable. 

The basic shift we're seeing is that rather than only following predetermined, programmed instructions, robots are now able to make their own judgments. 

Computers will increasingly be utilized for activities that involve decision-making, rather than just doing repetitive chores like applying formulae to numbers in a spreadsheet (which they currently excel at). 

This will inevitably contain choices affecting people's lives. 

The fact that Amazon is known to have fired human workers using AI algorithms. 

In such instance, it was discovered that robots handled the whole process of overseeing and firing employees. 

Beyond that, we are aware that AI has been used to the creation of weapons to produce tools that can decide whether to murder people (currently the UN states that it is unacceptable for them to be allowed to do this ). 

What about circumstances in which robots are forced to kill? 

How should a self-driving vehicle behave while deciding whether to collide with a pedestrian or a brick wall is a typical scenario used when debating AI ethics (potentially injuring its own driver). 

These are extreme instances, but they are pertinent to any ethical decisions you would need to make if you're in charge of deploying AI-powered, automated decision making at any firm. 

The most promising AI technologies all have extremely strong potential benefits, but if they are handled improperly, they might also have very strong potential drawbacks. 

In medical imaging, machine vision can identify malignant growths, but it may also be employed by authoritarian governments to monitor their population. 

It is now simpler than ever to communicate with robots, and natural language processing can scan social media postings for indications that someone could be suicidal or depressed. 

By pretending to be someone else, it may also be utilized to plan frauds and phishing assaults. 

And although we may use intelligent robots to investigate the ocean below or the universe, aid the disabled, and clean up the environment, they are also being created for use as lethal weapons. 

It's crucial that we consider where our own implementations of AI fall on this spectrum. 

Can we be certain that a concept like, for instance, the data-driven customer segmentation in our ice cream parlor would lie on the correct side of the moral divide? 

Hopefully no one reading this book is intending on using it to harm anybody. 

In such case, concerns about permission and privacy would be the main issues. 

It follows that it is unethical to gather, utilize, or distribute anyone's personal data without their consent if we recognize that individuals have a right to privacy. 

Therefore, getting agreement for any work we undertake with clients' private data is a crucial first step. 

Of course, this is now a legal necessity in many areas.

However, when it comes to the governance of your data or AI initiatives, ethics and law must both be handled separately. 

For instance, it's unlikely that Amazon's automatic termination of workers is against the law. 

On the other hand, it's pretty simple to construct a compelling case for it being immoral since it includes giving robots the authority to make decisions that may have a significant influence on people's lives. 

Furthermore, it's far from certain that the people involved ever agreed to give the robots this level of control over their lives. 

Even if you are just using computers to execute processes more effectively rather than to make such life-altering choices, caution must be taken. 

In 2018, IT contractor Ibrahim Diallo discovered he couldn't enter his Los Angeles workplace because his security permit had been suspended. 

After gaining entry, he was unable to enter into any of the systems he needed to use for his profession, and soon after that, security personnel showed up to take him out of the facility. 

Also terminated was his salary. 

His boss and any other senior employees were unaware of what had occurred, but after being compelled to work from home for three weeks, they discovered that an HR error had incorrectly indicated that he had been fired. 

At that moment, automatic systems started running and there was no way to manually override them. 

He had to quit the organization and look for employment elsewhere because of the stress and his supervisors' inability to take any action. 

What happens if workers are dismissed but not because of AI? 

According to some sources, IBM's staff headcount decreased by 25% between 2012 and 2019, and many of the tasks performed by individuals who departed were replaced by machines. 

There are undoubtedly compelling arguments that AI creates jobs rather than eliminates them for human workers. 

In addition, the types of jobs that it creates, such as those for engineers, scientists, data storytellers, and translators, are probably more satisfying than the jobs that are lost. 

According to the World Economic Forum, 85 million human jobs would have been replaced by automation by 2025. 

On the other hand, 97 million new jobs will be created within the same time period as a result of the rise in AI and other sophisticated, automated technologies. 

Does it imply that, in terms of replacing human labor, automation is not ethically problematic? Well, not really, I'd say! 

This implies that you must carefully examine how your own AI, automation, and data projects will affect your human workforce. 

• Is there a possibility that someone may develop redundancy? 
• If so, might they be given additional jobs in accordance with their talents and responsibilities, but with an emphasis on using the human traits that computers still lack, such as creativity, empathy, and communication? 
• Most importantly, is there a chance that they may lose importance regardless of whether you use your initiative? 
• If you didn't implement it, for instance, would it affect your company's ability to compete in the market, making it impossible for you to continue paying them? 

Further reason for worry comes from an AI technique known as a generative adversarial network (GAN), which enables the creation of uncannily realistic representations of actual individuals as well as text and language. 

This is what causes the "deepfake" pictures that have proliferated online. 

This has the potential to cause a great deal of harm to both individuals and society. 

For instance, it is very easy to find fake pornographic images of just about any celebrity online. 

It has also been used for political purposes to create images of politicians and world leaders saying or acting in ways that they would never do in real life. 

This may exacerbate the issues brought on by false information and the simplicity with which false or misleading information may be disseminated online, thereby undermining the democratic process. 

The generation of "synthetic data," and text-to-image systems, which essentially allow us to make pictures by describing what they look like, are just two of the many valid and totally ethical reasons a corporation would wish to employ GAN technology. 

The ethical risk is that they may also be used to produce anything that aids in the dissemination of incorrect information or results in the misrepresentation of specific persons or groups of people. 

Once again, if you use this technology, you need to be aware of the hazards and know where your beliefs fall in terms of ethics. 

From an ethical perspective, recommendation technology could appear like a secure use of AI, but there are also risks. 

We are given additional stuff to look at that is comparable to the content we have already looked at through systems that provide content, like Facebook's algorithms. 

This might result in "filter bubbles"; for instance, if you start reading the many conspiracy theory websites discussing topics like whether 9/11 was a "inside job" or the more current QAnon conspiracy, you will soon notice that similar items are starting to appear in your news feed. 

The same holds true whether you lean heavily to the left or right and often read news articles that support their viewpoints. 

The information you have accessible to support your own thoughts and beliefs may eventually become imbalanced if you only see tales that match that certain narrative. 

Again, this might have extremely negative effects on the democratic process by creating the echo-chamber effect. 

The need of transparency is still another problem. 

If you're going to place humans in a situation where robots control their lives, you should at the very least be able to describe how the machines make decisions. 

We've previously discussed the "black box" issue with AI, which arises from the fact that the algorithms may get so complicated that it is extremely challenging for humans to fully comprehend why they function in a certain way. 

This is made worse by the fact that some AI is purposefully opaque in order to deter easy duplication by those who sell it. 

Ethics are a personal matter that each individual must consider for himself, but you could discover that you agree with the majority of people who believe that actions that have an influence on people's lives should be explicable. 

It's interesting to note that some people think AR and VR might be useful in this situation by enabling us to study and question algorithmic operations in ways that aren't achievable by merely looking at computer code on a screen or even 2D representations of the operations. 

The OECD's Principles for AI include transparency as one of its primary needs (more on this below). 

Ethical issues are also brought up by AI's effects on the environment. 

According to estimates, the amount of carbon produced by training certain NLP machine learning models is equivalent to the yearly carbon emissions of 17 Americans. 

All that computing power uses quite a lot of energy. 

This is equivalent to around 283.048 kg (626,000 pounds) of CO2. 

Due to their offsetting methods, some AI providers, like Google, claim that they are carbon neutral, while others, like Microsoft, won't be until years or decades from now. 

Of course, AI has the ability to increase operational efficiency in many sectors, which may reduce the effect on the environment. 

For instance, both the network-wide strategies utilized by power and utility companies and the smart home thermostats are focused on utilizing energy more effectively and reducing emissions. 

There is no clear-cut answer to the question "Is AI good or bad for the environment?" because it depends on individual applications, which means that each application needs to be evaluated on its own merit for environmental impact. 

This is similar to the ethics of particular AI capabilities like facial recognition and natural language, and the implications on jobs. 

Given all the worries, it shouldn't be shocking that the AI business and larger society organizations are beginning to take the question of advanced technology ethics seriously. 

The Organization for Economic Co-operation and Development (OECD) has released its AI Principles, which include the following: 

AI development should be done in a way that respects the law, human rights, democracy, and diversity; it should be transparent to allow it to be understood and challenged; and organizations should be held responsible for the results of their AI initiatives. 

Google, too, has put forth its own standards of ethical AI usage. 

These include demands that AI apps be ethical, impartial, trustworthy, and considerate of users' privacy. 

Additionally, the business has said that it would seek to prevent hazardous or abusive implementations of AI. 

Additionally, Microsoft has a program called AI for Good that encourages the use of AI to address issues related to the environment, society, healthcare, and humanitarianism. 

The underuse of AI is my last point for this section, which I really believe needs thorough study. 

Simply said, there are numerous circumstances when avoiding AI is a bad idea from an ethical standpoint. 

• Doesn't that imply we have an ethical obligation to employ AI if there are problems in our business, society, or the world at large that might be solved with it but we choose not to - possibly because of other factors we've covered here, such the effect on the environment or transparency concerns? 

• Is it appropriate, for instance, to use facial recognition technology to locate someone who may be in danger even if they haven't given their consent to be followed in this way? 

Police have used it to locate persons who have gone missing in China and the UK. 

My suggestion is that any company intending to use AI take the step of establishing some kind of "ethical committee," similar to what Google and Microsoft have done. 

Of course, depending on the size of the company, your council's scope, size, and resources will vary. 

However, it's crucial to have someone in charge of taking into account all the concerns brought up here and how they affect our activities. 

~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram

References And Further Reading

1. Roff, HM and Moyes, R (2016) Meaningful Human Control, Artificial Intelligence and Autonomous Weapons, Briefing paper prepared for the Informal Meeting of Experts on Lethal Autonomous Weapons Systems, UN Convention on Certain Conventional Weapons, April, article36.org/wp-content/uploads/2016/04/MHC-AI-and-AWS-FINAL.pdf (archived at https://perma.cc/LE7C-TCDV)
2. Wakefield, J (2018) The man who was fired by a machine, BBC, 21 June, www.bbc.co.uk/news/technology-44561838 (archived at https://perma.cc/KWD2-XPGR)
3. Kande, M and Sönmez, M (2020) Don’t fear AI. It will lead to long-term job growth, WEF, 26 October, www.weforum.org/agenda/2020/10/dont-fear-ai-it-will-lead-to-long-term-job-growth/ (archived at https://perma.cc/LY4N-NCKM)
4. The Royal Society (2019) Explainable AI: the basics, November, royalsociety.org/-/media/policy/projects/explainable-ai/AI-and-interpretability-policy-briefing.pdf (archived at https://perma.cc/XXZ9-M27U)
5. Hao, K (2019) Training a single AI model can emit as much carbon as five cars in their lifetimes, MIT Technology Review, 6 June, www.technologyreview.com/2019/06/06/239031/training-a-single-ai-model-can-emit-as-much-carbon-as-five-cars-in-their-lifetimes/ (archived at https://perma.cc/AYN9-C8X9)
6. Najibi, A (2020) Racial discrimination in face recognition technology, SITN Harvard University, 24 October, sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/ (archived at https://perma.cc/F8TC-RPHW)
7. McDonald, H (2019) AI expert calls for end to UK use of ‘racially biased’ algorithms, Guardian, 12 December, www.theguardian.com/technology/2019/dec/12/ai-end-uk-use-racially-biased-algorithms-noel-sharkey (archived at https://perma.cc/WX8L-YEK8)
8. Dastin, J (2018) Amazon scraps secret AI recruiting tool that showed bias against women, Reuters, 11 October, www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G (archived at https://perma.cc/WYS6-R7CC)
9. Johnson, J (2021) Cyber crime: number of breaches and records exposed 2005–2020, Statista, 3 March, www.statista.com/statistics/273550/data-breaches-recorded-in-the-united-states-by-number-of-breaches-and-records-exposed/ (archived at https://perma.cc/BQ95-2YW2)
10. Palmer, D (2021) These new vulnerabilities put millions of IoT devices at risk, so patch now, ZDNet, 13 April, www.zdnet.com/article/these-new-vulnerabilities-millions-of-iot-devives-at-risk-so-patch-now/ (archived at https://perma.cc/RM6B-TSL3)