...
Ethical Issues Related to Artificial Intelligence
A presentation by:
Joel Castiglione, Robin Peskin, Cass Simmons, & Taylor Weida
University of Maryland Global Campus IFSM 304: Ethics in Information Technology Professor D. Hill 11MAY2021
Introduction:
Artificial Intelligence (AI) creates various Utilitarian ethical dilemmas
AI discrimination/bias
Privacy
Manipulation of human behavior
Impact OF THE Deepfake
As humanity has progressed through the digital age, consisting of the later half of the 20th Century into the 21st Century, Artificial Intelligences (AI) have grown and expanded from simple algorithmic functions found in a calculator to building search algorithms that can determine the next meal you crave. Advancements in this field are continually introducing new machines with the capacity to learn and perform cognitive tasks that previously only humans had the capability to complete. AI are becoming inseparable from humans in this digital age, and this is reflected through the increased integration of AI into business industries. With AI entering into business more and more, various consequence-driven ethical dilemmas have arisen that, if exploited, can lead to devastating affects resulting in discrimination, infringement of an individual’s right to privacy, manipulation of how people think, and the questionable trustworthiness of altered media through deepfakes. Understanding these utilitarian dilemmas can provide a benchmark for individuals to interpret what possible consequences exist by feeding information into the AI algorithms that exist in the digital age.
2
Artificial intelligence: Consequences of Discrimination and Bias
Does AI inadvertently discriminate against or show biases in favor of any group(s)?
“….replicates and embed the biases that already exist in our society.” (Pazzanese, 2020)
Societal impacts/Access to AI
Financial/monetary influence
Applying/hiring in the workforce
Locations/population
AI is seen by many to be impartial, but that is not the case as it can be prone to biases due to data inputs, algorithms, and other variables. This might lead to a system that “replicates and embed the biases that already exist in our society.” (Pazzanese, 2020)
AI does not possess the capacity to recognize when a pattern that brings positive results is also treating a particular group unfairly. This can lead to unforeseen negative consequences if not swiftly noticed and addressed, especially considering that the technology is becoming more pervasive in all aspects of daily life from the smart devices many own to our workplaces
AI has integrated itself into the day-to-day lives of countless individuals with most never noticing. Determining who gets approved for credit/loans and for how much, who gets the call for a job interview, or placed into a position, allocation of government subsidies (welfare, school placement), and even designating an area as “High Risk” for crime is all controlled in some way, shape, or form by AI (Smith, 2020).
3
Artificial intelligence: Discrimination and Bias (cont’d)
Human Creations
Biased Algorithm/Data
Facial Recognition (Apple/Microsoft/IBM)
“AI is not neutral: AI-based decisions are susceptible to inaccuracies, discriminatory outcomes, embedded or inserted bias.” (UNESCO, 2020)
Biased AI (Source: Smith,2020)
Artificial intelligence is programmed and created by human beings with their own sets of ethical beliefs and opinions. These mindsets are often programmed into the algorithms that in return gather/search/process the information based on potentially biased programming, which consequently can even lead to misidentifying information and groups of individuals.
“Labeling of data involves human discretion and can be subjective and discriminatory building from harmful biases, naivety, priorities and perspectives.” (Smith, 2020).
Amazon’s attempt at utilizing AI in its hiring process being a prominent example. Amazon expected their algorithm to maximize utility by selecting the five best resumes regardless of the number of resumes it was presented, but swiftly discovered that the AI strongly favored men. (Dastin, 2018) This error was due to the data sets it had received.
In 2018, researchers discovered that facial recognition software used by IBM, Microsoft, and Face++ were found to discriminate based on race and gender by primarily identifying white men and misidentifying the darker completions of both men and women (Palmer, 2018).
4
Artificial Intelligence: Privacy Issues
Need for lots of data
Third parties profit
Loss of Anonymity
(Batista, 2017)
Artificial intelligence is built to come up with the best possible solutions through correlating thousands of data points to achieve a result; the more data that is collected and processed, the more accurate the result ("Five essential capabilities”, 2019). This leads to privacy issues. When an individual agrees to the Terms of Service for a website, service, platform, etc., they agree to give their data. Sometimes an individual can adjust how much data they give, but no matter what, they are still sharing some data about themselves, even if it is your external IP and general geographic location. This isn’t inherently unethical since it is the trade of information in place of monetary costs, but the problem appears when consent is not given, requests to delete are denied, or when third parties get involved.
Sometimes upon agreeing to the Terms of Service, you also agree to allow your data to be shared, given, sold, or auctioned to third parties. The consequences that make this an ethical issue is that now you don’t know where your data is or who has access to it. For example, Google creates ‘profiles’ of it’s users and auctions that data to the bidders (Cyphers, 2020). This is probably why about 8 months ago when I was using Google’s search engine to shop for a shower curtain, and within five minutes I saw Amazon ads on Facebook for shower curtains.
Sometimes when signing up for a service, you have the choice to anonymize your data. But is anonymous really “anonymous”? In 2019 a news outlet was able to track President Trump using location data sent by one of his party member’s phones. By analyzing the location activity and other publicly accessible sources, the news team was able to track and identify the location and time of the President of the United States (Thompson, 2019). So, “Anonymous data” may not necessary be “anonymous”, especially when exposed to the power of AI.
5
Artificial Intelligence: Privacy Issues (CONT’d)
Smart City
Smart Home
Smart Wearables
The EU is anticipating the growth of large cities and realizes they will have a problem with infrastructure like waste management, security, energy, pollution, etc. To combat this issue and maximize utility, they plan to use emerging technologies such as AI. They plan on using lots of sensors to gather data such as noise monitoring, traffic patterns and speeds and correlating data to identifiable information like faces, body gait, and license plate numbers (Ryan, 2019). Consequently, the government will now have your facial recognition, license plate number to track where you go, how frequent you grocery shop, or shop at a cannabis or state store, how fast you drive, your predicted pattern to drive to these locations, etc. Where is the privacy in this?
And what of the old saying, “in the privacy of your own home”? A prime example of this would be Google’s Nest device. When it was first introduced, Google neglected to inform consumers about the onboard microphone and referred to this negligence as “an error” (Evangelho, 2019). Google has even gone so far as to suggest that one should inform guests in their own home that they risk being under constant surveillance if the home is equipped with Google Nest products (Holmes, 2019). Google and other tech companies have been far from forthcoming with precisely how much data they collect and what applications that data is applied to. A whistle blower revealed that the greatest consequence that product users faced was that Google was allowing employees to listen to sensitive data to improve its services; this was after discovering that Amazon indefinitely stores data that it gains in a similar manner from its Alexa line of products (Bote, 2019).
Smart wearables have often caused controversy over the exact methods they employ in their data collection and if their parent companies are being truthful in when the device is monitoring the users. When one uses a smartwatch to track their heart rate and other vitals tied to a workout app, is that Personal Health Information (PHI) stored securely, or is it more algorithmic data that we surrender to third parties to be used for future business advertisement targeting?
In a bid to optimize our lives, we risk our privacy and anonymity being taken from us. Think about how much data will be collected about yourself living in a Smart City within a Smart Home using Smart Wearables; what data feeds will they use to gather this information? What private third parties would make that information accessible to the government? And what extra data will they collect with that shared information? How will your anonymity be kept in the aggregation and correlation of this data?
6
Artificial Intelligence: Manipulation of human behavior
of amazon sales influenced by ai algorithm
Netflix guides 80% of viewers to their next show using ai
ai increases addictive behaviors to social media
People who excessively use social media negatively impacted with smaller reward center in brain
inhibited sense of danger
An often-overlooked issue concerning AI is its ability to manipulate human behavior. AI possesses an increasingly pervasive influence on our decision-making. On Amazon, one-third of all choices made are influenced by their algorithm. This percentage skyrockets to 80% of all viewing activity on Netflix where their proprietary algorithm has guided customers to their next show or movie (Hosanagar, 2019). Social media sites, online video games, and even dating apps are all actively vying for the consumer’s time and attention, willing to be exploitative of human nature to gain any utilitarian business advantage through retention and lengthened user engagement. At times, these methods could result in unintended consequences that harm the users. Such consequences include how social media addiction could possibly affect anywhere from 5 to 10% of the US population. Additionally, studies have shown that users under 21 who excessively use social media undergo structural changes in their brains that result in a smaller reward center (Ricci, 2018). Our ever-growing dependence on AI presents another issue: the false assumption that human oversight is unnecessary for certain technologies. Driverless cars are one such example as many assume Tesla vehicles are fully capable of driving themselves without human assistance. Unfortunately, consequence of this has already resulted in multiple deaths with one example having no human present in the driver’s seat when the vehicle crashed (Kolodny, 2021).
7
Artificial intelligence: Consequences of Deepfake
Impact to personal or professional standing
Political influence (Toews, 2020)
Loss of public trust
Loss of opportunity
Creation of disinformation
Deepfake of Russian President Putin (Source: Bloomberg, 2018).
Artificial Intelligence Synthesized Content, colloquially referred to as Deepfake, is the false attribution of speech or activities to an individual via the creation and publication of realistic and malicious videos, photos, or recordings of an individual acting or speaking in a manner that the actual person never would. The intent behind the creator of the deepfake is to cause personal or professional harm, sway public opinion, or to generate illicit financial gain. There is a high potential for misuse by individuals, nations, and everything in between. World leaders who hold strong influence and power have been subject to deepfake AI, such as the President of the United States, Barack Obama and President of the Russian Federation, Vladimir Putin (Toews, 2020). Deepfake AI is becoming increasingly prevalent and harder to detect, and no current recourse for Deepfake AI victims exist as legal liability is hard to assess against a backdrop of the First Amendment’s right to freedom of expression and speech.
Potential fallout of misattributing speech to an adversary [Image from video] (Bloomberg, 2018).
8
Artificial intelligence: consequences of Deepfake
Reputational Impact
Potential for leverage actions (blackmail, voice-spoofing, etc.)
Financial Loss (Stupp, 2019)
Emotional distress
In what is recognized as the first AI cybercrime, A UK CEO’s voice was
impersonated to initiate a financial transfer. Company loss was $243,000.
Deep fakes are used to negatively influence people’s thoughts and beliefs about, or to take advantage of, the victim.
This can be achieved through disinformation campaigns using manipulated material in an attack that currently only affects women, like deepfake pornography (Engler, 2020).
These attacks can have long lasting effects on the victim’s personal reputation, livelihood and potentially even their mental health.
Cyber-criminals mimicked the voice of the CEO of a UK energy firm, convincing him he was speaking to the CEO of the German parent firm. The deepfake CEO requested an immediate funds transfer to a third-party supplier. After the transfer was made it came to light that the entire conversation was faked by AI voice-spoofing. The money was quickly moved to other banks, and no suspects have been identified. This event is considered by many to be the first AI cyber-crime (Stupp, 2019).
9
Summary/REVIEW
Artificial Intelligence (AI) creates various Utilitarian ethical dilemmas
AI bias due to data inputs
Privacy threatened by third-party data sharing
human behavior influenced by ai recommendations
Deepfakes impacting trust in global politics and finance
We discussed a few of the utilitarian ethical dilemmas surrounding Artificial Intelligence to include AI discrimination and bias, matters of privacy, the affects on human behavior, and the impacts of deepfakes.
10
Conclusion
Artificial Intelligence (AI) Is Extremely influential
AI creates various Utilitarian ethical dilemmas
Consequence-driven results could negatively impact society
Laws should be put in place and enforced to prevent ethical erosion
Humans and ai can best coexist when ethical principles are applied
In conclusion, Artificial Intelligence is revolutionizing the world and could potentially become the most influential human innovation in history. As with the Industrial Revolution that came before it, there is a varying number of ethical dilemmas raised by the utilitarian implementation of this new technological advancement. Unlike previous Ages in human history however, the risk of further embedding discrimination in our society, erasure of anonymity through massive privacy degradation, the threat of innate human behavior being manipulated and the rise of deepfakes to cause personal or professional harm is too great to ignore. Doing so will result in consequence-driven results that could negatively impact both individuals and society for decades to come and beyond. Instead, regulations and laws are necessary to ensure that entities employing AI act in a manner that maximizes the health and well-being of everyone equally while also respecting their privacy and autonomy. Addressing these matters will help ensure that humanity and AI can coexist in harmony.
11
References:
Batista, E. (2017, March 20). The problem with Anonymous feedback. Ed Batista. https://www.edbatista.com/2017/03/the-problem-with-anonymous-feedback.html
Bloomberg Quicktake (2018, September 27). It’s Getting Harder to Spot Aa Deep Fake Video [Video]. Youtube. https://www.youtube.com/watch?v=gLoI9hAX9dw.
Bote, J. (2019, July 12). Google workers are eavesdropping on your private conversations via its smart speakers. USA Today. https://www.usatoday.com/story/tech/2019/07/11/google-home-smart-speakers-employees-listen-conversations/1702205001/.
Cyphers, B. (2020, July 10). Google says it doesn’t 'Sell' your data. Here’s how the company shares, monetizes, and exploits it. Electronic Frontier Foundation. https://www.eff.org/deeplinks/2020/03/google-says-it-doesnt-sell-your-data-heres-how-company-shares-monetizes-and
Dastin, J. (2018, October 10). Amazon scraps secret AI recruiting tool that showed bias against women. Reuters. https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G.
Engler, A. (2020, May 6). Fighting deepfakes when detection fails. Brookings. https://www.brookings.edu/research/fighting-deepfakes-when-detection-fails/.
Evangelho, J. (2019, February 20). Google: It Was An 'Error' Not Telling Anyone About Secret Nest Microphone. Forbes. https://www.forbes.com/sites/jasonevangelho/2019/02/20/google-it-was-an-error-not-telling-anyone-about-secret-nest-microphone/?sh=6f8b033c4165.
Five essential capabilities: Automated machine learning. (2019, January 28). Ironside - Business Analytics. Data Science. Information Management. https://www.ironsidegroup.com/2018/06/06/five-essential-capabilities-automated-machine-learning/
Holmes, A. (2019, October 16). Google exec says Nest owners should probably warn their guests that their conversations are being recorded. Business Insider. https://www.businessinsider.com/google-exec-nest-owners-should-tell-guests-theyre-being-recorded-2019-10.
Hosanagar, K. B. P. R. V. N. A. (2019, March 25). How Algorithms Impact Our Decisions. Who Made That Decision: You or an Algorithm? https://knowledge.wharton.upenn.edu/article/algorithms-decision-making/.
Johnson, D.G., & Diakopoulos, N. (2021). Computing Ethics: What To Do About Deepfakes. Communications of the ACM, 64(3), 33-35. https://doi-org.ezproxy.umgc.edu/10.1145/3447255
Kolodny, L. (2021, April 19). 'No one was driving' in Tesla crash that killed two men in Spring, Texas, report says. CNBC. https://www.cnbc.com/2021/04/18/no-one-was-driving-in-tesla-crash-that-killed-two-men-in-spring-texas-report.html.
Palmer, A. (2018, February 12). Study finds popular face ID systems may have racial bias. Daily Mail Online. https://www.dailymail.co.uk/sciencetech/article-5382979/Study-finds-popular-face-ID-systems-racial-bias.html.
Pazzanese, C. (2020, December 4). Ethical concerns mount as AI takes bigger decision-making role. Harvard Gazette. https://news.harvard.edu/gazette/story/2020/10/ethical-concerns-mount-as-ai-takes-bigger-decision-making-role/.
Ricci, J. (2018, June 28). The Growing Case for Social Media Addiction. The California State University. https://www2.calstate.edu/csu-system/news/Pages/Social-Media-Addiction.aspx.
Ryan, M., & Gregory, A. (2019). Ethics of Using Smart City AI and Big Data: The Case of Four Large European Cities. ORBIT Journal, 2(2). https:// doi.org /10.29297/orbit.v2i2.110
Stupp, Catherine (2019, August 30). Fraudsters Used AI to Mimic CEO’s voice in unusual cybercrime case. Wall Street Journal. https://www.wsj.com/articles/fraudsters-use-ai-to-mimic-ceos-voice-in-unusual-cybercrime-case-11567157402?utm_campaign=the_algorithm.unpaid.engagement&utm_source=hs_email&utm_medium=email&utm_content=76535820&_hsenc=p2ANqtz-_HP64YUhkAXhxAYRrsaGBy1NcusfB8RBDpoNn3Ie2_GdN85NJIwUWmmCPzGwLg7LHx2su7J9hasgZSzGFdAUzZXbJpVg&_hsmi=76535820
Smith, G., & Rustagi, I. (2020). Mitigating Bias in Artificial Intelligence: An Equity Fluent Leadership Playbook . University of California. https://haas.berkeley.edu/wp-content/uploads/UCB_Playbook_R10_V2_spreads2.pdf.
Thompson, S., & Warzel, C. (2019, December 22). How to track President Trump. The New York Times - Breaking News, US News, World News and Videos. https://www.nytimes.com/interactive/2019/12/20/opinion/location-data-national-security.html.
Toews, R. (2020, May 25). Deepfakes Are Going To Wreak Havoc On Society. We Are Not Prepared. Forbes. Retrieved from: https://www.forbes.com/sites/robtoews/2020/05/25/deepfakes-are-going-to-wreak-havoc-on-society-we-are-not-prepared/
UNESCO. (2020, October 2). Artificial Intelligence: examples of ethical dilemmas. UNESCO. https://en.unesco.org/artificial-intelligence/ethics/cases.
12