|
Final Assignment for JOUR 3751 at UMN
Introduction Before the rise of digital media in journalism, humans determined what was newsworthy using their prior knowledge and investigative research, constrained by their physical and mental capacity to write articles before a deadline, and based on their experience during their lifetime or the shared experiences and values of a newsroom. As digital media became pervasive during the 20th and 21st centuries, news has benefited from the inputs of computer algorithms that can write, research, and moderate at least as efficiently as the human journalists on which the computer scientists have based their designs. Three aspects of news production have changed: article writing has been enhanced by machine learning algorithms, computer science has expanded the capacity of news organizations to leverage research, and finally, technology has improved the efficiency and civility of audience-provided feedback to the news organization. The importance of these technological changes has been measured economically through standard metrics like unique visitors, illustrated by a case study of Le Monde, below. Though the value of other improvements, such as greater gender inclusivity and civility, have been difficult to measure they are equally important to society. Finally, optimization of basic journalistic processes by digital technology led to more efficient newsrooms that produce more content so that they can be relevant to a wider audience. News production before digital media Before digital media, local beat writers would cover one sports team or one sport in their area, speaking to athletes and taking notes on a pad or recording device that required transcription, both of which required additional time to reorganize before someone could begin to write a story. News organizations were constrained by their human resources, for example, “AI tools can help journalists tell new kinds of stories that were previously too resource-impractical or technically out of reach” (Hansen et al., 2017). Physical proximity to the subjects of their stories also constrained the organizations. The financial costs of travel, in addition to the travel time that was not conducive to productive writing, have been mitigated by technology. Once someone had written a story during the era before digital media, it would be given to another person in the newsroom who focused on production tasks, such as layout, bylines, a headline, and images. These copy editors were experts at writing headlines based on a formula of engagement so that people would buy and subscribe to newspapers. They had a knack for selecting captivating images, too. Unfortunately, “Copy editors have been sacrificed more than any other newsroom category” (Beaujon, 2013). Algorithms are most effective when replacing human activities that are based on a “knack” because those human skills tend to use factors that can be captured and built into a computer model. While much of the attention on artificial intelligence in the newsroom has been focused on the bias within the algorithms, it is important to understand the context of bias in newsrooms before digital media, which continues today. News production has been overwhelmingly white, and predominantly male, and remains this way today according to a Women’s Media Center analysis of the “Five Big Sunday Shows” in 2020 in which “findings confirm many years of research… as ample evidence shows, the news is fundamentally male” (Byerly, 2021). This article also refers to a study by Gaye Tuchman from over 40 years ago that quantified gender representation in news organizations as overwhelmingly white male. Perspectives offered by the news before digital media tended to follow the ideology of the writers and production staff, who were white, and though the statistics indicate that minority and female representation has not changed with the demographic changes in the United States, technology is beginning to offer insights about inclusivity and more in-depth research from more points of view. I believe that the first step toward being inclusive is being open to feedback. We are imperfect beings who benefit from listening to others’ viewpoints. In the past, the audience of journalism wrote letters to the newspaper. Large newsrooms invested in mailroom headcount to receive and sort, then route the opinions and feedback of their readers to the parts of the organization that produced a particular section. I experienced a world that required a visit to the post office or a visit from the postman, plus a few days for a letter to travel to the news organization, then a day or more for the letter to be open, read, and sorted. The feedback loop between audience and news producer required a week or longer. Reacting to feedback about a story was a slow process in most circumstances. Also, rude or nasty letters could be sorted into a separate basket that was routed differently, to avoid offending the writer and/or the producers. The feedback loop has been one of the biggest changes in digital media. Effect of technology in the age of digital media Many major news organizations maintain a robot writer on staff. “Cyborg… accounts for an estimated one-third of the content published by Bloomberg News” (Contributor, 2019) and “Bertie is part of a broader focus on using artificial intelligence to make publishing more efficient for Forbes staff” (Willens, 2019). Those are just two examples of many algorithms that are writing the content for news publishers. These companies maintain a robot writer, not a staff of robot writers, to supplement the production of their human writers. “The AP estimated that it’s freed up 20 percent of reporters’ time spent covering corporate earnings and that AI is also moving the needle on accuracy” (Moses, 2017). I think the coolest news writing algorithm is GPT-3 that, or who (depending on bias toward machine learning algorithms), writes for The Guardian. Following a recent article that GPT-3 titled “A robot wrote this entire article. Are you scared yet, human?” the editor provided a brief recap of the process they used: “prompts were written by The Guardian… GPT-3 produced eight different outputs… Each was unique, interesting, and advanced a different argument.” (Editor, 2020) Like they would have done for a human writer, the editor picked “the best parts of each… Editing GPT-3’s op-end was no different to editing a human op-ed. Overall, it took less time to edit than many human op-eds.” Efficiency has been a huge benefit of algorithms, as well as several others that will be discussed later. These Robot Writers reduce the time needed to write the article, as the AP noted, and they are more accurate because computers are parameter-based technology that struggles with intuitive decision-making. They stay on topic. The second writing technology reduces the time to write headlines and choose images, “reinforcement learning can also be applied to optimize publishing; for example, to help choose the best headlines or thumbnails for a particular story.” (Marconi, 2020). The methodology of good headlines was based on the experience of copy editors, which has been recreated in machine learning algorithms that are reactive to feedback and produce more options, more quickly (Clark, 2019). A case study of the UK press association reported, “journalists have developed templates for particular topics and use automation to create multiple variations” (Marconi, 2020) which are presented to editors as suggested headlines, images, and entire stories, just like GPT-3. After discussing two good examples of technological efficiency, let’s examine how algorithms can affect inclusivity and gender bias. In a predominantly white male news organization, connecting stories to female and minority points of view can be difficult. Algorithms are being used by the Financial Times to track personal pronoun usage. “As reporters write their piece, the bot will alert them of any imbalance in gender ratios” and the same technology is being used to monitor images to ensure there is gender equity in coverage, “based on ‘research showing a positive correlation between stories including quotes of women and higher rates of engagement with female readers’” (Marconi, 2020). Human writers are the arbiter of this algorithmic feedback, choosing to adjust their story based on suggestions, or not, but the correlation to engagement translates to marketing revenue and higher-quality work, so it is likely that the technology will continue to be deployed. The computer processing power and databases required to support these on-demand suggestions are immense, but the infrastructure exists already. Known as neural networks because their complexity is like “the way neurons are wired in the brain” (Hutson, 2021), these algorithms can browse massive datasets in milliseconds to find connections within the data, connections based on templates and prompts created by newsrooms. “Content and news organizations are making increasing use of AI systems to uncover data from multiple sources and automatically summarize them into articles or supporting research for those articles” (Schmelzer, 2019). A case study of the 2015 French election coverage by Le Monde explains how the news organization beat a major competitor “by having more published stories online” (Marconi, 2020) increasing page views, which is also happening in the Real Estate- and Sports-writing industries (Duncan, 2020). Quality content that is “automatically generated provides a net value for us – and our readers” has increased conversions of new paying subscribers (Kalim, 2021) just like coverage of corporate earnings. Finally, the most obvious new technology is the Comment feature in digital media. This pervasive and ubiquitous technology enables an audience to give feedback immediately, not a week later, and anyone can access their feedback. Algorithms are being used to moderate the comments, group comments for human review, and score comments for civility. The New York Times (NYT) uses a tool called Moderator that has allowed them to open more stories for comments (Etim, 2016; Salganik & Lee, 2020). Not just an efficiency tool, the “main goal is to create a safe space for discussions” (Kovalyova, 2021) that drives engagement with readers. Internally, these moderator tools are used to fact-check reporters, too, because both processes depend on accuracy. Collaboration, a strength of networked journalism, improves the entire news-gathering process from story research and writing to the comments posted by audiences. Personal Reflection and Conclusion The ethical obligations of organizations that use algorithms have been altered by digital media. Bylines include references to the robot reporters based on suggested headlines and images. More news content is being written as audience participation escalates through comments. I appreciate the benefits of efficiency and inclusivity in the form of more articles that offer a broader perspective, even if the writer is a machine, or a machine influences the writer. Egalitarian systems that allow everyone to post feedback improve the user experience of journalism by eliminating the delays created by snail mail and by forcing news organizations to address alternative opinions that are posted online. Many of these systems exist in “black boxes” though, so only a privileged few understand the inputs and outputs of the algorithms. Transparency is being prioritized by organizations like NYT and my hope is that this open-source approach continues. I think it will continue because the business models measure the financial benefits through audience engagement. I believe that algorithms will save us in the end, because they can overcome the limits of the human mind. Only an algorithm can easily track our use of personal pronouns in our articles, and mention when we’re off track. News writing, research, and audience participation have been changed by digital media. The slow pace of writing in the past is exponentially faster using algorithms. News producers’ dependence on “the use of algorithms to automatically generate news from structured data has shaken up the journalism industry” (Graefe, 2016). Disruption of old enterprises is a good thing, though we need to carefully monitor ethical concerns. News is being optimized by computer technology, but it is also being optimized by greater audience participation, the inclusion of more gender and minority perspectives, and giving human writers more time to reflect and revise. References
0 Comments
No Black Boxes (Procedural Transparency): Refers to the transparency necessary to understand and measure the effects of outcomes and outputs. This requires proper documentation of the algorithm as well as proper retention policies to ensure the system designs, database schemas, and major product decisions have been clearly defined in detail. Ethical concerns caused by inconclusive evidence for algorithmic outputs are compounded by challenges with tracing the decision process of machine intelligence but being able to reference documentation that explains the decision tree of the algorithm ensures responsibility for agency through the attribution of action to designers, circumstances, and/or machine intelligence.
Predictability and Explainability: To properly attribute agency of the algorithmic outputs, the forementioned transparency is a useful tool, but additionally, the designers and technical teams that support the development of the algorithm need to ensure that their design has consistently produced the expected outputs along with the evidence to explain the logic the algorithm used to make its decisions. Most notably in unsupervised machine learning algorithms, it is difficult to differentiate between a designed outcome of an algorithm vs. an unexpected outcome based on what the algorithm learned independently. As Safiya Noble noted, algorithms are “automated decisions” that we must trust, most easily accomplished by inspection of the decision-making process to measure its predictability. Aggressive Collection of Feedback (user and technical): Solicit input from the most varied group of stakeholders possible. Similar to a Works Council in Germany that includes a representative from every facet of the business – from janitorial to management to senior leadership – consideration of any new updates to the algorithm must be authorized by large, broadly representative groups of users and stakeholders. Users often have more detailed product knowledge than designers, like the poet in this week’s video, Joy Buolamwini, who noticed the problem with facial recognition technology and black skin. Stakeholders often possess a broader awareness of extra-contextual factors that may influence the algorithm. Both groups must be frequently consulted by the teams that design, build, and maintain the algorithm. Ongoing maintenance: Systematic collection of feedback from users will identify problems with the current design and opportunities for improvement based on new technology and the wisdom gained through experience. Sometimes, an algorithm will do harm to its users, so it is essential to continuously solicit feedback, analyze reporting produced by the algorithm, and aggressively monitor performance trends. Observations or reports of harm to users should generate an immediate maintenance response from the design team and a fair evaluation of the potential need to suspend the algorithm based on the scale, and correctability, of the harm. Since most algorithms are more complex than a checkers game (which we learned has been process-mapped to prevent algorithms from losing a match) the proper balance of accuracy and efficiency, a core value of algorithms, will almost always require more updates in the future. Data Security: Require a minimum standard for the protection of data created by the algorithm, stored by technical teams, and used by additional parties. To accomplish this goal, using the best hardware, firmware, and software will be expected, but most importantly, using the right technology to minimize the risk of a data breach, and any potential misuse of the data. Privacy is a right in some countries, protecting individual autonomy by limiting the unauthorized collection of data, as well as improper use. Most algorithms create more data than a human mind can monitor, so implementing the right tools – building them if necessary for special requirements – will protect data throughout the end-to-end process. Processes should be designed to protect data and all relevant work teams should be fully trained in data security. Midterm Assignment for CI 4311W at UMN
Introduction One of the greatest accomplishments of Artificial Intelligence has been Machine Learning Algorithms (MLAs). They are not the same as average, everyday algorithms with a ‘little a’. The biggest difference is that MLAs are semi-autonomous at least, sometimes entirely autonomous, and if either condition is true, that creates ethical complexities related to individual responsibility, attribution of virtues, vices, and beneficiaries, as well as the transparency necessary to understand and measure the effects of outcomes and outputs. Furthermore, the autonomy of MLAs puts them into a unique class with ambiguous rights and duties. To unravel some of these complexities, at least enough to engage in a basic ethical discussion about them, we need to examine a machine-oriented definition of intelligence and rationality. We will need to reflect on problems as old as the Wizard of Oz, the movie through which the world witnessed the transformation of moving pictures from black-and-white to color; a radical change. The transformation of algorithms to MLAs has a comparable scale, i.e., from two colors to infinite colors. A simple and well-known example of an MLA is the spam filter in email; semi-autonomous because we click buttons in unsolicited emails to help train it. On the other hand, the “k-means clustering” MLA learns and operates unsupervised, independently discovering answers to questions about observations that have not occurred in a human mind yet. These MLAs exceed the brain’s capacity to process information. I will use this type of algorithm as a reference point because I have worked directly with it to build global fraud prevention programs. On several occasions, my brain was the first step of the validation process to ensure that a newly coded k-means clustering analytic tool was accurate and useful. I am neither mathematically inclined, nor scientific – my input was the external reference point. It was important that I did not fully understand the mechanisms of MLAs, and that remains true today. Ethics of Evidence My human brain was the bulwark to filter inconclusive evidence, an epistemic ethical concern. Algorithms “encourage the practice of apophenia: ‘seeing patterns where none actually exist, simply because massive quantities of data can offer connections that radiate in all directions’ (Boyd and Crawford 2012, 668)” (Tsarnados et al., 2021). In ‘The ethics of algorithms: key problems and solutions’ a team of researchers provides two additional types of evidentiary concerns, both more ethically concerning: inconclusive evidence and misguided evidence. The latter creates risks to justice and fairness because neither the algorithm’s designers nor the MLA has been well prepared to assess the social context that the outputs affect. In addition, once the MLA has been implemented, the technical team will move to a new project or new jobs, creating a risk of context transfer bias should the new owners of the algorithm apply it to a novel problem. Ethical concerns with inconclusive evidence are compounded by challenges with tracing the decision process of machine intelligence. This transparency problem is another significant ethical concern related to the traceability of machines’ learning and subsequent decision-making. The ability to understand algorithms is limited to a relatively small audience, often affluent and educated, furthering social inequality as MLAs become more prolific and integrated into our basic tools for daily living. Documentation is too much for users, too difficult to produce by business, and slows innovation in return for ubiquitous benefits. A related problem is attributing responsibility for the actions of MLAs. “The technical complexity and dynamism of ML algorithms make them prone to concerns of ‘agency laundering’: a moral wrong which consists in distancing oneself from morally suspect actions, regardless of whether those actions were intended or not, by blaming the algorithm (Rubel et al. 2019)” (Tsarnados et al., 2021). The compound effect of a small group of knowledgeable owners of machine learning algorithms and the general difficulty in determining when an algorithmic output is the result of human design or independent and unsupervised machine learning leads to conflicts of moral duties. These conflicts are created by the black box in which the algorithm exists and the utilitarian difficulties created by challenges predicting consequences when only a small group of experts control the machine, or the machine makes decisions by itself based on the rules it has written, which are based on the learning in which it has engaged independently. It is important to note that transparency is a human problem too because most people cannot explain the ’what, how, and why’ for every action. Presuming that this human limitation is accurate, we would be treating MLAs unfairly by expecting it in their domain. Intelligence, Rationality, and the Ethics of Agency Two common philosophical definitions of Intelligence apply to our understanding of MLAs. First, intelligence is the capacity to achieve complex goals, and second, intelligence does the right thing at the right time. “The teleological framework of multiple goals, within which humans move about balancing reason and emotion, brings us back to Aristotle’s view of ethics as the pursuit of the good life through a well-orchestrated ethos. Aristotle highlights the capacity to listen as one of the key competencies of ethical-practical reasoning, phronesis, which he also calls a ‘sense’” (Holst, 2021). In my work validating MLAs’ k-means clustering analysis, it was what had been professionally labeled my “Spidey sense” for uncovering complex, well-hidden webs of fraud that was being compared to the outputs of the intelligent, unsupervised machine. It is important to note that my intuition was less transparent than the outputs of the machine that was being built to replace me. Predictability and explainability are both important aspects of virtuous decisions. ‘An FDA for Algorithms’ highlights “the analogy between complex algorithms and complex drugs. With respect to the operation of many drugs, the precise mechanisms by which they produce their benefits and harms are not well understood. The same will soon be true of the most important (and potentially dangerous) future algorithms” (Tutt, 2016). The prospect of becoming a full ethical agent is a human problem that applies to MLAs. For Plato, it is a two-part problem, first, “in his dialogues, the crucial capacity for rational thinking is called ‘logon didonai’ which means to give an account of something” (Holst, 2021). We have explored that problem of transparency already. The second part, developed further and adopted by Aristotle, includes desire and emotion – both mastery and the use of those irrational components of reason. How does one apply this second part of being a full ethical agent to a machine? This is the existential dilemma faced by the Tin Man in the Wizard of Oz: he does not have a heart and thus, cannot experience emotion. One can imagine a machine designed for this purpose, but the complexity of designing a problem-solving machine like an MLA with these components embedded seems unfeasible and impractical. “In contrast to reinforcement learning methods, both supervised and unsupervised learning methods are… poorly suited methods for the raising of ethical machines” (Kaas, 2021) Two topics mentioned in different articles in the book “Machine Law, Ethics, and Morality in the Age of Artificial Intelligence” can be combined to propose a solution. In a game called Cake or Death, the subject will be presented with two options, both of which lead to ethically appropriate rewards – bake a cake or kill three people. There is a third option, too, after which the program loops back to the original decision – ask a virtual entity for advice. Additional, more complex, games were outlined in Kaas’ article, but at the core of each one was a nudge of the machine to learn to ask for help when making decisions, and they present all decisions as benefitting from this ethical process through a reward mechanism. With this approach, ethical MLAs would be designed, implemented, and maintained for use by other MLAs that need this specific kind of advice, using common technological procedures like today’s APIs that already request and receive the high volumes of data needed to support frequent ethics checks. Conclusions and Implications These are just a few of the ethical concerns with machine learning algorithms. One simple solution addresses concerns with the ethical categories that are the broad focus of this paper (Rights, Utilitarian Good, Fairness, Common Good, and Virtue): ongoing maintenance. The technical teams that build these machines will transition to new roles and projects. Contexts in which these machines are used will change over time. The mathematics and science that created MLAs will produce new discoveries that will incrementally, sometimes fundamentally, change the technology. For these reasons, maintenance is an ethical requirement for the builders and owners of MLAs, even when these roles are taken over by the MLAs. Maintenance enables input from the participants, users, and social contracts. The environment in which technology develops is naturally iterative so the maintenance will not be a burden to development. Also, using the ethical guidelines of pharmaceuticals may inform the ethics of intelligent machine learning algorithms. AlphaGo and Deep Blue were successful in their domains because they used probability theory refined through deep self-reflection achieved through the machines’ observations of the games they played against themselves. A machine playing a game against itself appears to be an independent entity, which is why they call that machine learning process unsupervised. But the games these two MLAs play have minimal ethical concerns compared to the medical advice, fraud prevention, and language processing that the more well-integrated MLAs will produce. Finally, after his thorough discussion of the Black Box and Tin Man problems with artificial intelligence, he ends optimistically, and I would like to echo his thoughts as I conclude. “Due to a mismatch of our rational and emotional nature, many of our human ways of acting and thinking are flawed with incoherent, contradictory, and biased elements that still stand in the way of realizing fully ethical rationality. Should the future development of AI become truly ethical, it will also help us become more ethical.” (Holst, 2021) Because it is most likely that one day, we will rely on the outputs of MLAs in most aspects of our lives, assessing the ethics of our design decisions is an essential component of the MLA design process today. Successful, independent learning machines may help us overcome some of the human challenges that we all experience every day, most notably the struggle between emotional desire and our rational intelligence. It is my belief that these two constitutional aspects of humanity are not only the result of natural processes outside of our control but also the result of the learning processes that we engage throughout our lives. Because our machines have achieved the threshold of unsupervised learning, like their original makers, these infinite decisions from which we learn can be traced back to the black-and-white of ethically good or bad. References
|
AuthorStudent of Education, English, and Learning Technology at UMN. Archives
May 2022
Categories
All
|
RSS Feed