1.3 Computer Science and the Future of Society

Developing Foundational Technologies

We’ve seen how foundational technologies like artificial intelligence, algorithms, and mathematical models enable important applications in data science, computational science, and information science.

As noted previously, artificial intelligence (AI) is the development of computer functions to perform tasks, such as visual perception and decision-making processes, that usually are performed by human intelligence. AI refers to a subfield of CS that is interested in solving problems that require the application of machine learning to human cognitive processes to achieve goals. AI research seeks to develop algorithm architectures that can make progress toward solving problems. One such example is image recognition, or the problem of identifying objects in an image. This problem is quite difficult for programmers to solve using traditional programming methods. Imagine having to define very precise rules or instructions that could identify an object in an image regardless of its position, size, lighting conditions, or perspective. As humans, we have an intuitive sense of the qualities of an object. However, representing this human intelligence in a machine requiring strict rules or instructions is a much harder task. AI methods for image recognition involve designing algorithm architectures that can generalize across all the possible ways that an object can appear in an image.

Recent approaches to AI for image recognition draw on a family of methods called neural networks instead of having programmers craft rules or instructions by hand to form an algorithm. In humans, the neural network is a complex network in the human brain that consists of neurons, or nerve cells, connected by synapses that send messages and electrical signals to all parts of the body, enabling us to think, move, feel, and function.

In computer science, a neural network (Figure 1.9) is an AI algorithm architecture that emphasizes connections between artificial nerve cells whose behavior and values change in response to stimulus or input. These neural networks are not defined by individual neurons but by the combination of all the neurons in the network. Typically, artificial neurons are arranged in a hierarchy that aims to capture the structure of an image. Although the first level of neurons might respond to individual pixels, later levels of artificial neurons might respond in aggregate to the arrangement of several artificial neurons in the preceding layer. This is similar to how the human visual system responds to edges at the lower levels, then responds in aggregate to the specific arrangement of several edges in later levels, and ultimately identifies these aggregated arrangements as objects.

Figure 1.9 (a) An artificial neural network consists of three key layers: the input layer, where raw data enters the system; the hidden layer, where information is processed and patterns are identified; and the output layer, where results are presented. (b) A natural neural network, such as those in the human body, mirrors this structure. The input layer represents sensory receptors, like those in the retina. The hidden layer corresponds to the synapse, where partial processing of the sensory data occurs. Finally, the output layer represents the information sent to the brain for final processing and interpretation. (credit a: modification of "Neural network example" by Wiso/Wikipedia, Public Domain; credit b: modification of work from Psychology 2e. attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

The idea of neural networks, however, is not as new as it might seem. Artificial neural networks were first imagined in the mid-1900s alongside contemporary research efforts in the cognitive sciences. The ideas of multilayered, hierarchical networks of neurons and the mathematical optimization methods for learning were all there, but these early efforts were limited by the computational processing power available at the time. In addition, the large datasets that drive neural network learning were not nearly as available as they are today with the Internet. Developments in foundational technologies such as computer architecture and computer networks paved the way for the more recent developments in neural network technologies. The broad area of computer systems investigates these architectures and networks that enable new algorithms and software. Without these technologies, neural networks would not be nearly as popular and revolutionary as they are today. Yet the relationship between computer systems and AI development is not one-directional. Today, computer scientists are using neural networks to help design new, more efficient computer systems. The development of foundational computer technologies not only creates opportunities for direct and indirect applications, but also supports the development of other computer technologies.

Just as we saw how technological fixes embodied a powerful belief about the relationship between computer solutions and social good, a similar cultural belief exists about the relationship between foundational technologies and their social values. The belief that technologies are inherently neutral and that it is the people using technology who ultimately make it “good” or “bad” is considered social determination of technology.

Today’s neural networks are designed to identify patterns and reproduce existing data. It is widely accepted that many big datasets can encode social preferences and values, particularly when the data is collected from users on the Internet. A social determination of technology accepts this explanation of AI bias and leaves the design of AI algorithms and techniques as neutral: the bias in an AI system is attributed to the social values of the data rather than the design of the AI algorithms. Critics of social determination point out that the way AI algorithms learn from big data represents a social value, one that encodes a default preference for reproducing the biases inherent in big data. This applies whether the AI application is about fair housing, medical imaging, ad targeting, drone strikes, or another topic. This is an issue that computer scientists must consider as they practice responsible computing and strive to ensure that data is gathered and handled as ethically as possible.

Evaluating Negative Consequences of Technology

Today’s AI technologies work by reproducing existing patterns rather than imagining radically different futures. As much as neural networks are inspired by the human brain, it would be a stretch to suggest that AI systems have any semblance of general intelligence. Though these systems might be quite effective at identifying lettuce plants from weed plants in an image, their capacity for humanlike intelligence is limited by design. A neural network learns to recognize similar patterns that appear across millions or billions of sample images and represent these patterns with millions or billions of numbers. Mathematical optimization methods are used to choose the numeric values that best encode correlations across the sample images. However, current approaches lack a deeper, conceptual representation of objects. One criticism of very large neural networks is that there are often more numeric values than there are sample images—the network can effectively memorize the details of a million sample images by encoding them in a billion numbers. Many of today’s neural networks recognize objects in images not by relying on some intrinsic idea or concept of objects but by memorizing every single configuration of edges as they appear in the sample images.

This limitation can lead to peculiar outcomes for image recognition systems. Often, neural network approaches for image recognition have certain examples of images where objects are misidentified in unusual ways: a person’s face might be recognized in a piece of toast or in a bunch of clouds in the sky. In these examples, the pattern of edges might coincidentally trigger the neural network values so that it misidentifies objects. These are among the more human-understandable examples; there are many other odd situations that are less explainable. An adversarial attack is a sample input (e.g., an image) that is designed to cause a system to behave problematically. Researchers have found that even tweaking the color of just a single point in an image can cause a chain reaction in the neural network, leading it to severely misidentify objects. The adversary can choose the color of the point in such a way as to almost entirely control the output of some neural networks: changing a single specific point in an image of a dog might cause the system to recognize the object as a car, airplane, human, or almost anything that the adversary so desires. Moreover, these adversarial attacks can often be engineered to cause the neural network to report extremely high confidence in its wrong answers. Self-driving cars that use neural networks for image recognition can be at risk of real-world adversarial attacks when specially designed stickers are placed on signs that cause the system to recognize a red light as a green light (Figure 1.10). By studying adversarial attacks, researchers can design neural networks that are more robust and resilient to these attacks.

Figure 1.10 (a) Autopilot functions in self-driving cars generally identify roads and lanes using artificial intelligence to “see” road markings. (b) Researchers were able to trick these cars into seeing new lanes by using as few as three small stickers, to confuse the neural networks and force the cars to change lanes. (attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

In general, research is an important part of computer science. Through research, computer scientists analyze ways that technology can be used and gain insight and answers to address issues and improve various aspects of society. Research enables computer scientists to make advancements like the design of new algorithms, development of new hardware and software, and applications for emerging technologies such as AI.

One important use of research is to investigate adversarial attacks to gather answers needed for computer scientists to improve foundational technologies by evaluating the negative consequences of technologies. Computer technologies offer a unique medium for learning things (not just learning computer science), connecting with each other, and enhancing the lives of people all around the world. Yet, in each of these examples, we also raised concerns about how these technologies unfolded and affected people’s lives in both positive and negative ways. While we can rarely, if ever, paint any one technology as purely “good” or “bad,” computer scientists are interested in studying questions around how technologies are designed to center social values. Social scientists are not solely responsible for answering questions about technology, but computer scientists can also contribute important knowledge and methods toward understanding computer technologies.

Designing Technologies for Social Good

Computer science can advance social good by benefiting many people in many different areas, including public health, agricultural sustainability, climate sustainability, and education.

Computer technologies accelerate medical treatments for public and personal health from initial research and development to clinical trials to large-scale production and distribution. In January 2020, Chinese officials posted the genetic sequence of the coronavirus SARS-CoV-2. This helped pharmaceutical companies to begin developing potential vaccines for the virus at a significantly faster rate than for any other virus in the past (Figure 1.11).

Figure 1.11 The SARS-CoV-2 outbreak that began in 2020 displayed how quickly computer science could be harnessed by governments, medical facilities, and scientists to decode the virus, develop treatments, and distribute vaccinations around the world. What would have been a very difficult feat to manage manually was simplified through the use of algorithms and computer technology. (credit left: modification of "COVID-19 vaccines" by Agência Brasília/Flickr, CC BY 2.0; credit center: modification of "T04" by Sarah Taylor/Flickr, CC BY 2.0; credit right: modification of “Back2School Brigade prepares families for upcoming school year” by Thomas Karol/DVIDS, Public Domain)

Computational science enables the miracles of modern medicine. Viral sequences can be digitized and rapidly shared between researchers across the world via the Internet. Computer algorithms and models can simulate the human immune system responses to particular treatments within hours rather than years. The first treatments can then be produced at a small scale using computer-engineered cells in less than a month from the initial sequencing. To ensure the treatments are safe and effective, clinical trials are held at disease transmission “hot spots” predicted using data science methods drawing on data aggregated and monitored from across the world. Once a treatment is proven safe and effective, it is mass-produced with the help of computer-controlled robots and automated assembly lines. Algorithms manage the inventory supply and demand and control the transportation of treatments on trucks and planes guided by computer navigation systems. Web apps and services notify people throughout the process.

Yet the use of computer technology throughout modern medicine is anything but politically neutral. Computers, algorithms, and mathematical models solve the problems that their creators wish to solve and encode the assumptions of their target populations. Supply and demand data for the data models are determined by various factors, at least partly in response to the money and relationships between countries that control the technology, the Global North, and countries that don’t, the Global South. Within local communities, the uptake of medical treatments is often inequitable, reflecting and reinforcing historical inequities and disparities in public health. Computer technology alone often doesn’t address these issues. In fact, without people thinking about these issues, computer technologies can often amplify disparities. Consider datasets, which can be biased if they overrepresent or underrepresent specific groups of people. If decisions are made on the basis of biased data, people in the groups that are not represented fairly may receive inequitable treatment. For example, if a local government agency is working with a biased dataset, political leaders may make decisions that result in certain citizens receiving inadequate funding or services. This is an example of why responsible computing, which we will cover in Chapter 14 Cyber Resources Qualities and Cyber Computing Governance, is so important.

These problematic histories are not only aggravated in medicine and public health, but also reflected in housing. Redlining refers to the inequitable access to basic public services based on residents’ neighborhoods and communities, which includes the practice of withholding financial services from areas with a large underrepresented population. In the United States, these communities reflect the histories of racial segregation and racial wealth inequalities. Fair housing laws are intended to prevent property owners from discriminating against buyers or renters because of race, color, ability, national origin, and other protected classes. But computer technologies also present new kinds of challenges. Microtargeted ads on social media platforms contribute to not only political polarization, but also discrimination in housing. This can be a particular problem when combined with redlining. Even if the ad targeting is not explicitly designed to discriminate, microtargeted ads can still reinforce historical redlining by incorporating data such as zip codes or neighborhoods. This may result in digital redlining, which is the practice of using technology, such as targeted ads, to promote discrimination. In 2021, a Facebook user filed a class-action lawsuit that argued nine companies in the Washington, D.C., area deliberately excluded people over the age of 50 from seeing their advertisements for housing because they wanted to attract younger people to live in their apartments.¹⁶ This is an example of an issue in technology that should be addressed by responsible computing with an emphasis on ethical behavior.

With good intentions and attention to personal biases, technologies can be designed for social good. For example, a hypothetical algorithm for fair housing could evenly distribute new housing to people across protected classes and marginalized identities, such as older populations. Of course, algorithmic control and automated decision-making is challenged to consider the underlying conditions behind social problems. Still, algorithms can be important tools to enable us to distribute outcomes more fairly from a statistical perspective, and this can be an important step in addressing the larger societal systems and inequities that produce social problems.

As part of responsible computing, computer scientists must be aware of technological fix, which refers to the idea that technologies can solve social problems, but is now often used to critique blind faith in technological solutions to human problems. Unless the process is handled responsibly, the “fix” may cause more problems than it resolves. When considering how to address social and political problems, computer scientists must take care to ensure that they select the appropriate technology to address specific problems.

To address social problems and advance social good, recall that human-centered computing emphasizes people rather than technologies in the design of computer solutions. A human-centered approach to fair housing might begin by centering local communities directly affected by redlining. Rather than replacing or disrupting the people and organizations already working on a problem, a human-centered approach would center them in the design process as experts. A human-centered approach requires that the designer ask why they are not already working with people in the community impacted by their work.