Rhetorical Strategies in Machine Learning
Student Name
Dept. of Rhetoric & Writing Studies, San Diego State University
RWS 305W: Writing in Various Settings
Professor Centanni
10 October 2019
Rhetorical Strategies in Machine Learning[LC2]
Being part of a specific community entails sharing the same values as other members in the community, as well as oftentimes following the same ways of communicating those values. This is notably the case within the field of machine learning, which is a part of the larger field of computer science and software. Machine learning, which happens to be a rapidly growing yet somewhat controversial part of software, is essentially the use of algorithms and patterns to train a machine to do or recognize certain things without explicit instructions. In this field, which encompasses many new technologies like facial recognition and object detection, the typical communication patterns become clear after reading through various scholarly journals on the topic. Within the field of machine learning, authors employ statistics, popular references, acronyms, and other strategies in order to appeal to their intended audiences and show their authority within a discourse community[LC3] .
Before examining how rhetorical strategies were used by the particular authors under analysis, it is important to observe that common target audiences in this field includes students and members of the field of computer science who may not know much about machine learning. This becomes evident when each of the writings makes an effort to initially define what machine learning and artificial intelligence are, and what purposes they may help serve. For example, in the article “Philosophy and Machine Learning,” Canadian philosopher Paul Thagard [LC4] (1990) describes the aim of artificial intelligence as “getting computers to perform tasks that require intelligence when done by people” (p. 261). By defining the aim of artificial intelligence in easily understandable terms, Thagard makes it clear that his writing is meant to reach readers who may not currently know much about machine learning or about programming at all. If readers were expected to have an abundance of experience, definitions would not be necessary. This is only one sign of the audience, but being able to identify the intended readers clears up why the authors choose to use the various rhetorical strategies they later employ in the interest of persuasion.
One such [LC5] strategy is citing statistics. Researchers Barrington et al. (2012), in “Game-Powered Machine Learning,” cite statistics early and often in their discussion on machine learning. By the second paragraph, the authors’ use of statistics shows the reader the power of machine learning when used to create song recommendations for music listeners on different streaming platforms. This is shown when Barrington et al. (2012) notes “after 10 [years] of effort by up to 50 full time musicologists, less than 1 million songs have been manually annotated, representing less than 5% of the current iTunes catalog” (p. 1611). Here, Barrington et al. use statistics to convince the reader that humans who manually annotate songs lack the efficiency of computer systems that do the same.[LC6] By comparing human and machine data entry through illustrative statistics, Barrington et al. are able to appeal to logos to show the immense difference in efficiency between human annotating and the use of machine learning to accomplish the same task. This is an important signifier of a field norm because readers in this field do not just expect conjecture; they want to know that concrete data supports assertions. While statistics can certainly be manipulated if used incorrectly, this field values those that are put into an appropriate context.[LC7]
While statistics are often utilized in illustrating widespread effects of machine learning to audiences, it is also important that the authors are able to connect the subject to readers on a personal level as well[LC8] . A prime example of this is shown in the article “Machine Learning” by Dellot and Balaram (2018). Early on in the piece, the authors show their effort to connect to the reader by naming popular shows in which dystopian futures become the norm as a somewhat direct result of machine learning. They connect with readers by noting “popular culture is again dominated by tales of machines gone rogue, from Ex Machina to Black Mirror” (Dellot & Balaram, 2018, p. 44). [LC9] Here, it is clear that the authors are trying to reel in the interest of the audience by almost immediately connecting the subject to television shows they might be familiar with or may have heard of. By doing this, they appeal to pathos in order to keep the reader intrigued through imagery and emotional connection. It creates an immediate sense of relevance in a reader, which shows that the argument is not just about dry claims and statistics, but also about relatable concepts. Appealing to pathos through nostalgic experience or real-life examples is an evident and seemingly necessary strategy frequently used within the field of machine learning to create a connection with the reader.
Though creating a personal connection to the audience is important to stimulate the interest of the readers in the widespread and statistic-filled subject of machine learning, it is imperative that the reader is able to see the author as an authority within the discourse community of computer science as a whole. One of the most commonplace ways authors are able to show their authority in computer programming and belonging to the discourse community of machine learning is through the use of acronyms. A simple example of this is shown when Thagard (1990) almost immediately shortens artificial intelligence to AI (p. 261), but a more complex illustration occurs when Barrington et al. (2012) shortens “Gaussian mixture model” and “dynamic texture mixture” to GMM and DTM, respectively (p. 1614). Here, the authors of both texts employ acronyms to abbreviate terms that they will use regularly in efforts to demonstrate their understanding and experience with the subject of machine learning. To be sure, knowing that AI means “artificial intelligence” is not insider knowledge; however, by meeting the norm of employing the common acronyms on a frequent basis, these authors implant the idea that they do have further insider knowledge, which makes them authorities in the field. There are many terms within the computer science and machine learning field that are long and specific, and these acronyms both aid in creating a sense of belonging within the field and in allowing people to reference concepts without adding confusing, lengthy definitions. By abbreviating these terms, the authors are able to appeal to ethos in the subject of machine learning while also allowing themselves to use the terms continually throughout the writings without tiring them out.
Although [LC10] writers in the discourse community of machine learning do well in displaying their experience within the field, one strategy that is notably absent from the field is authors directly referring to the experience. In other words, they commonly show their experience without talking about it. Whereas this may be a more suitable strategy to employ in other fields of study, it is not often used when discussing machine learning, likely since the field is such a new one that is rapidly growing and changing every day. Claiming 10 years of experience in a field is not always particularly impressive, but few people have been able to study machine learning much longer, since the earliest texts come from about 30 years ago, when the field itself was still budding. With a field that is ever-changing and continually improving, it becomes difficult to truly be an expert or to bring up experience with machine learning as past experiences can quickly grow outdated as new technology develops. With these thoughts in mind, it is clear why appeals to ethos are less often used in writings on machine learning.
The rhetorical strategies within the field of machine learning begin to reveal themselves after reading through scholarly journals on the topic. Strategies used to appeal to pathos, logos, and ethos assume a pattern and are utilized by members of the discourse community to show their belonging to the field. Writers are oftentimes found using statistics in order to appeal to logos, as well as using media and imagery to emotionally connect the reader to machine learning by appealing to pathos. Using acronyms to appeal to ethos become more apparent since this rhetorical appeal is found few and far between within this discourse community. As the field of machine learning continues to advance and grow, the rhetorical strategies used by members of its discourse community are one element that seem to maintain a steady pattern[LC11] .
References[LC12]
Barrington, L., Turnbull, D., & Lanckriet, G. (2012). Game-powered machine learning. Proceedings of the National Academy of Sciences of the United States of America, 109(17), pp. 6411–6416. https://doi.org/10.1073/pnas.1014748109
Dellot, B., & Balaram. B. (2018). Machine learning. RSA Journal, 164(3[5575]), pp. 44-47. https://doi.org/10.2307/26798354
Thagard, P. (1990). Philosophy and machine learning. Canadian Journal of Philosophy, 20(2), pp. 261–276. https://doi.org/10.1080/00455091.1990.10717218
[LC1]Note the perfect formatting on this page, including the bold title, a space after, and the full Department and Institution. Always spell out dates rather than using the numerical form.
[LC2]In APA 7th edition, the title recurs, in bold, on page 2.
[LC3]The introduction begins broad, by discussing “communities” in general. It then narrows down to the specific topic of the paper. By the end, the thesis statement clarifies exactly where this paper will go.
[LC4]Note that APA in-text citations include author, year, and page number. They can either come in this format, called a “narrative citation,” or they can be entirely within parentheses at the end of a sentence. See page 4 for an example of a parenthetical citation.
[LC5]This sentence uses the phrase “one such” to call back to the final sentence of the preceding paragraph. This is a strong transition, and it creates flow between ideas.
[LC6]After every quotation or paraphrase, you should spend one sentence interpreting the meaning of what you just cited. Do not assume your reader will understand it as you do.
[LC7]Note how roughly half of the paragraph is devoted to interpretation, explanation, and analysis of the specific quote displayed. Avoid generalizations after giving quotes/paraphrases and, instead, constantly refer back to what you just showed your reader.
[LC8]Note that every single body paragraph begins with a specific topic sentence that introduces the main point that will be proven. Do not give general ideas that don’t guide your reader.
[LC9]Here is the parenthetical citation. Since the sentence did not begin with the author name(s), it came in parentheses.
[LC10]This paragraph should have evidence. It is difficult to find evidence of something that does not appear in your field, but an alternative would be to show how your field circumvents the strategy you identify.
For example, this writer could have shown one of the authors displaying their experience without calling attention to it and noted what was missing in the evidence.
[LC11]Conclude by “zooming out” once more to the concept of discourse communities, at large.
[LC12]Again, notice the perfect formatting here. The writer puts “References” in bold (not Works Cited or Bibliography). They cite all author names as APA instructs, place the dates appropriately, capitalize titles as APA prefers, and italicizes where necessary.
DOI entries can often be found in the databases, but notice how they ALL look the exact same. If your DOI does not begin https://doi.org/, then it is incorrect. For more help, see Formatting Your References List.
Lastly, the only reference entries that have URL instead of DOI should be those that a) do not have a DOI, and b) have a static, consistent URL (in other words, it does not require a login). See the above resource for more information.