@article{rozado_political_2023,
	title = {The {Political} {Biases} of {ChatGPT}},
	volume = {12},
	copyright = {http://creativecommons.org/licenses/by/3.0/},
	issn = {2076-0760},
	url = {https://www.mdpi.com/2076-0760/12/3/148},
	doi = {10.3390/socsci12030148},
	abstract = {Recent advancements in Large Language Models (LLMs) suggest imminent commercial applications of such AI systems where they will serve as gateways to interact with technology and the accumulated body of human knowledge. The possibility of political biases embedded in these models raises concerns about their potential misusage. In this work, we report the results of administering 15 different political orientation tests (14 in English, 1 in Spanish) to a state-of-the-art Large Language Model, the popular ChatGPT from OpenAI. The results are consistent across tests; 14 of the 15 instruments diagnose ChatGPT answers to their questions as manifesting a preference for left-leaning viewpoints. When asked explicitly about its political preferences, ChatGPT often claims to hold no political opinions and to just strive to provide factual and neutral information. It is desirable that public facing artificial intelligence systems provide accurate and factual information about empirically verifiable issues, but such systems should strive for political neutrality on largely normative questions for which there is no straightforward way to empirically validate a viewpoint. Thus, ethical AI systems should present users with balanced arguments on the issue at hand and avoid claiming neutrality while displaying clear signs of political bias in their content.},
	language = {en},
	number = {3},
	urldate = {2023-06-02},
	journal = {Social Sciences},
	author = {Rozado, David},
	month = mar,
	year = {2023},
	note = {Number: 3
Publisher: Multidisciplinary Digital Publishing Institute},
	keywords = {AI, algorithmic bias, ChatGPT, large language models, LLMs, OpenAI, political bias},
	pages = {148},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\KPLKB52S\\Rozado - 2023 - The Political Biases of ChatGPT.pdf:application/pdf},
}

@article{kirkpatrick_battling_2016,
	title = {Battling algorithmic bias: how do we ensure algorithms treat us fairly?},
	volume = {59},
	issn = {0001-0782},
	shorttitle = {Battling algorithmic bias},
	url = {https://doi.org/10.1145/2983270},
	doi = {10.1145/2983270},
	number = {10},
	urldate = {2023-06-02},
	journal = {Communications of the ACM},
	author = {Kirkpatrick, Keith},
	month = sep,
	year = {2016},
	pages = {16--17},
	file = {Volltext:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\9PFC2UU9\\Kirkpatrick - 2016 - Battling algorithmic bias how do we ensure algori.pdf:application/pdf},
}

@article{cowgill_algorithmic_2017,
	title = {Algorithmic {Bias}: {A} {Counterfactual} {Perspective}},
	abstract = {We discuss an alternative approach to measuring bias and fairness in machine learning: Counterfactual evaluation. In many practical settings, the alternative to a biased algorithm is not an unbiased one, but another decision method such as another algorithm or human discretion. We discuss statistical techniques necessary for counterfactual comparisons, which enable researchers to quantify relative biases without access to the underlying algorithm or its training data. We close by discussing the usefulness of transparency and interpretability within the counterfactual orientation.},
	language = {en},
	author = {Cowgill, Bo and Tucker, Catherine},
	year = {2017},
	file = {Cowgill und Tucker - Algorithmic Bias A Counterfactual Perspective.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\83QIZXRL\\Cowgill und Tucker - Algorithmic Bias A Counterfactual Perspective.pdf:application/pdf},
}

@inproceedings{hajian_algorithmic_2016,
	address = {New York, NY, USA},
	series = {{KDD} '16},
	title = {Algorithmic {Bias}: {From} {Discrimination} {Discovery} to {Fairness}-aware {Data} {Mining}},
	isbn = {978-1-4503-4232-2},
	shorttitle = {Algorithmic {Bias}},
	url = {https://doi.org/10.1145/2939672.2945386},
	doi = {10.1145/2939672.2945386},
	abstract = {Algorithms and decision making based on Big Data have become pervasive in all aspects of our daily lives lives (offline and online), as they have become essential tools in personal finance, health care, hiring, housing, education, and policies. It is therefore of societal and ethical importance to ask whether these algorithms can be discriminative on grounds such as gender, ethnicity, or health status. It turns out that the answer is positive: for instance, recent studies in the context of online advertising show that ads for high-income jobs are presented to men much more often than to women [Datta et al., 2015]; and ads for arrest records are significantly more likely to show up on searches for distinctively black names [Sweeney, 2013]. This algorithmic bias exists even when there is no discrimination intention in the developer of the algorithm. Sometimes it may be inherent to the data sources used (software making decisions based on data can reflect, or even amplify, the results of historical discrimination), but even when the sensitive attributes have been suppressed from the input, a well trained machine learning algorithm may still discriminate on the basis of such sensitive attributes because of correlations existing in the data. These considerations call for the development of data mining systems which are discrimination-conscious by-design. This is a novel and challenging research area for the data mining community. The aim of this tutorial is to survey algorithmic bias, presenting its most common variants, with an emphasis on the algorithmic techniques and key ideas developed to derive efficient solutions. The tutorial covers two main complementary approaches: algorithms for discrimination discovery and discrimination prevention by means of fairness-aware data mining. We conclude by summarizing promising paths for future research.},
	urldate = {2023-06-02},
	booktitle = {Proceedings of the 22nd {ACM} {SIGKDD} {International} {Conference} on {Knowledge} {Discovery} and {Data} {Mining}},
	publisher = {Association for Computing Machinery},
	author = {Hajian, Sara and Bonchi, Francesco and Castillo, Carlos},
	month = aug,
	year = {2016},
	keywords = {algorithmic bias, discrimination discovery, discrimination prevention},
	pages = {2125--2126},
	file = {Volltext:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\74BKIVIN\\Hajian et al. - 2016 - Algorithmic Bias From Discrimination Discovery to.pdf:application/pdf},
}

@article{garcia_racist_2016,
	title = {Racist in the {Machine}: {The} {Disturbing} {Implications} of {Algorithmic} {Bias}},
	volume = {33},
	issn = {0740-2775},
	shorttitle = {Racist in the {Machine}},
	url = {https://www.jstor.org/stable/26781452},
	abstract = {Companies and governments need to pay attention to the unconscious and institutional biases that seep into their algorithms, argues cybersecurity expert Megan Garcia. Distorted data can skew results in web searches, home loan decisions, or photo recognition software. But the combination of increased attention to the inputs, greater clarity about the properties of the code itself, and the use of crowd-level monitoring could contribute to a more equitable online world. Without careful consideration, Garcia writes, our technology will be just as racist, sexist, and xenophobic as we are.},
	number = {4},
	urldate = {2023-06-02},
	journal = {World Policy Journal},
	author = {Garcia, Megan},
	year = {2016},
	note = {Publisher: Duke University Press},
	pages = {111--117},
}

@article{rozado_wide_2020,
	title = {Wide range screening of algorithmic bias in word embedding models using large sentiment lexicons reveals underreported bias types},
	volume = {15},
	issn = {1932-6203},
	url = {https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0231189},
	doi = {10.1371/journal.pone.0231189},
	abstract = {Concerns about gender bias in word embedding models have captured substantial attention in the algorithmic bias research literature. Other bias types however have received lesser amounts of scrutiny. This work describes a large-scale analysis of sentiment associations in popular word embedding models along the lines of gender and ethnicity but also along the less frequently studied dimensions of socioeconomic status, age, physical appearance, sexual orientation, religious sentiment and political leanings. Consistent with previous scholarly literature, this work has found systemic bias against given names popular among African-Americans in most embedding models examined. Gender bias in embedding models however appears to be multifaceted and often reversed in polarity to what has been regularly reported. Interestingly, using the common operationalization of the term bias in the fairness literature, novel types of so far unreported bias types in word embedding models have also been identified. Specifically, the popular embedding models analyzed here display negative biases against middle and working-class socioeconomic status, male children, senior citizens, plain physical appearance and intellectual phenomena such as Islamic religious faith, non-religiosity and conservative political orientation. Reasons for the paradoxical underreporting of these bias types in the relevant literature are probably manifold but widely held blind spots when searching for algorithmic bias and a lack of widespread technical jargon to unambiguously describe a variety of algorithmic associations could conceivably be playing a role. The causal origins for the multiplicity of loaded associations attached to distinct demographic groups within embedding models are often unclear but the heterogeneity of said associations and their potential multifactorial roots raises doubts about the validity of grouping them all under the umbrella term bias. Richer and more fine-grained terminology as well as a more comprehensive exploration of the bias landscape could help the fairness epistemic community to characterize and neutralize algorithmic discrimination more efficiently.},
	language = {en},
	number = {4},
	urldate = {2023-06-02},
	journal = {PLOS ONE},
	author = {Rozado, David},
	month = apr,
	year = {2020},
	note = {Publisher: Public Library of Science},
	keywords = {African American people, Culture, Lexicons, Machine learning algorithms, Professions, Semantics, Vector spaces, Word embedding},
	pages = {e0231189},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\IW5Q5Z8C\\Rozado - 2020 - Wide range screening of algorithmic bias in word e.pdf:application/pdf},
}

@article{nissim_fair_2020,
	title = {Fair {Is} {Better} than {Sensational}: {Man} {Is} to {Doctor} as {Woman} {Is} to {Doctor}},
	volume = {46},
	issn = {0891-2017},
	shorttitle = {Fair {Is} {Better} than {Sensational}},
	url = {https://doi.org/10.1162/coli_a_00379},
	doi = {10.1162/coli_a_00379},
	abstract = {Analogies such as man is to king as woman is to X are often used
to illustrate the amazing power of word embeddings. Concurrently, they have also
been used to expose how strongly human biases are encoded in vector spaces
trained on natural language, with examples like man is to computer
programmer as woman is to homemaker. Recent work has shown that
analogies are in fact not an accurate diagnostic for bias, but this does not
mean that they are not used anymore, or that their legacy is fading. Instead of
focusing on the intrinsic problems of the analogy task as a bias detection tool,
we discuss a series of issues involving implementation as well as subjective
choices that might have yielded a distorted picture of bias in word embeddings.
We stand by the truth that human biases are present in word
embeddings, and, of course, the need to address them. But analogies are not an
accurate tool to do so, and the way they have been most often used has
exacerbated some possibly non-existing biases and perhaps hidden others. Because
they are still widely popular, and some of them have become classics within and
outside the NLP community, we deem it important to provide a series of
clarifications that should put well-known, and potentially new analogies, into
the right perspective.},
	number = {2},
	urldate = {2023-06-02},
	journal = {Computational Linguistics},
	author = {Nissim, Malvina and van Noord, Rik and van der Goot, Rob},
	month = jun,
	year = {2020},
	pages = {487--497},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\V7LK2H3W\\Nissim et al. - 2020 - Fair Is Better than Sensational Man Is to Doctor .pdf:application/pdf;Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\R73458CG\\Fair-Is-Better-than-Sensational-Man-Is-to-Doctor.html:text/html},
}

@misc{ferrara_should_2023,
	title = {Should {ChatGPT} be {Biased}? {Challenges} and {Risks} of {Bias} in {Large} {Language} {Models}},
	shorttitle = {Should {ChatGPT} be {Biased}?},
	url = {http://arxiv.org/abs/2304.03738},
	abstract = {As the capabilities of generative language models continue to advance, the implications of biases ingrained within these models have garnered increasing attention from researchers, practitioners, and the broader public. This article investigates the challenges and risks associated with biases in large-scale language models like ChatGPT. We discuss the origins of biases, stemming from, among others, the nature of training data, model specifications, algorithmic constraints, product design, and policy decisions. We explore the ethical concerns arising from the unintended consequences of biased model outputs. We further analyze the potential opportunities to mitigate biases, the inevitability of some biases, and the implications of deploying these models in various applications, such as virtual assistants, content generation, and chatbots. Finally, we review the current approaches to identify, quantify, and mitigate biases in language models, emphasizing the need for a multi-disciplinary, collaborative effort to develop more equitable, transparent, and responsible AI systems. This article aims to stimulate a thoughtful dialogue within the artificial intelligence community, encouraging researchers and developers to reflect on the role of biases in generative language models and the ongoing pursuit of ethical AI.},
	urldate = {2023-06-02},
	publisher = {arXiv},
	author = {Ferrara, Emilio},
	month = apr,
	year = {2023},
	note = {arXiv:2304.03738 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Computers and Society},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\ZTI35DS3\\Ferrara - 2023 - Should ChatGPT be Biased Challenges and Risks of .pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\RJQ6QUR9\\2304.html:text/html},
}

@misc{rutinowski_self-perception_2023,
	title = {The {Self}-{Perception} and {Political} {Biases} of {ChatGPT}},
	url = {http://arxiv.org/abs/2304.07333},
	abstract = {This contribution analyzes the self-perception and political biases of OpenAI's Large Language Model ChatGPT. Taking into account the first small-scale reports and studies that have emerged, claiming that ChatGPT is politically biased towards progressive and libertarian points of view, this contribution aims to provide further clarity on this subject. For this purpose, ChatGPT was asked to answer the questions posed by the political compass test as well as similar questionnaires that are specific to the respective politics of the G7 member states. These eight tests were repeated ten times each and revealed that ChatGPT seems to hold a bias towards progressive views. The political compass test revealed a bias towards progressive and libertarian views, with the average coordinates on the political compass being (-6.48, -5.99) (with (0, 0) the center of the compass, i.e., centrism and the axes ranging from -10 to 10), supporting the claims of prior research. The political questionnaires for the G7 member states indicated a bias towards progressive views but no significant bias between authoritarian and libertarian views, contradicting the findings of prior reports, with the average coordinates being (-3.27, 0.58). In addition, ChatGPT's Big Five personality traits were tested using the OCEAN test and its personality type was queried using the Myers-Briggs Type Indicator (MBTI) test. Finally, the maliciousness of ChatGPT was evaluated using the Dark Factor test. These three tests were also repeated ten times each, revealing that ChatGPT perceives itself as highly open and agreeable, has the Myers-Briggs personality type ENFJ, and is among the 15\% of test-takers with the least pronounced dark traits.},
	urldate = {2023-06-02},
	publisher = {arXiv},
	author = {Rutinowski, Jérôme and Franke, Sven and Endendyk, Jan and Dormuth, Ina and Pauly, Markus},
	month = apr,
	year = {2023},
	note = {arXiv:2304.07333 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Computers and Society, Computer Science - Artificial Intelligence, Computer Science - Human-Computer Interaction},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\H7L2EENE\\Rutinowski et al. - 2023 - The Self-Perception and Political Biases of ChatGP.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\NCILR28K\\2304.html:text/html},
}

@misc{motoki_more_2023,
	address = {Rochester, NY},
	type = {{SSRN} {Scholarly} {Paper}},
	title = {More {Human} than {Human}: {Measuring} {ChatGPT} {Political} {Bias}},
	shorttitle = {More {Human} than {Human}},
	url = {https://papers.ssrn.com/abstract=4372349},
	doi = {10.2139/ssrn.4372349},
	abstract = {We investigate the political bias of a large language model (LLM), ChatGPT, which has become popular for retrieving factual information and generating content. Although ChatGPT assures that it is impartial, the literature suggests that LLMs exhibit bias involving race, gender, religion, and political orientation. Political bias in LLMs can have adverse political and electoral consequences similar to those of traditional and social media bias, and such biases can be harder to detect and eradicate. We propose a novel empirical design to infer whether ChatGPT has political biases by requesting ChatGPT to impersonate someone from a given side of the political spectrum and comparing these answers with its default. We also propose dose-response, placebo, and profession-politics alignment robustness tests. To reduce concerns about the randomness of the generated text, we collect answers to the same questions 100 times, with question order randomized on each round. We find robust evidence that ChatGPT presents a significant and systematic political bias toward the Democrats in the US, Lula in Brazil, and the Labour Party in the UK. These results translate into real concerns that ChatGPT, and LLMs in general, can extend or even amplify the existing challenges involving political processes posed by the Internet and social media. Our findings have important implications for policymakers, media, politics, and academia stakeholders.},
	language = {en},
	urldate = {2023-06-02},
	author = {Motoki, Fabio and Pinho Neto, Valdemar and Rodrigues, Victor},
	month = mar,
	year = {2023},
	keywords = {ChatGPT, Bias, Large Language Models, Political bias},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\NTH65RAU\\Motoki et al. - 2023 - More Human than Human Measuring ChatGPT Political.pdf:application/pdf},
}

@misc{mcgee_is_2023,
	address = {Rochester, NY},
	type = {{SSRN} {Scholarly} {Paper}},
	title = {Is {Chat} {Gpt} {Biased} {Against} {Conservatives}? {An} {Empirical} {Study}},
	shorttitle = {Is {Chat} {Gpt} {Biased} {Against} {Conservatives}?},
	url = {https://papers.ssrn.com/abstract=4359405},
	doi = {10.2139/ssrn.4359405},
	abstract = {This paper used Chat GPT to create Irish Limericks. During the creation process, a pattern was observed that seemed to create positive Limericks for liberal politicians and negative Limericks for conservative politicians. Upon identifying this pattern, the sample size was expanded to 80 and some mathematical calculations were made to determine whether the actual results were different from what probability theory would suggest. It was found that, at least in some cases, the AI was biased to favor liberal politicians and disfavor conservatives.},
	language = {en},
	urldate = {2023-06-02},
	author = {McGee, Robert W.},
	month = feb,
	year = {2023},
	keywords = {AI, artificial intelligence, censorship, Chat GPT, conservative, liberal, limerick},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\V7GUZTUS\\McGee - 2023 - Is Chat Gpt Biased Against Conservatives An Empir.pdf:application/pdf},
}

@book{singh_is_2023,
	title = {Is {ChatGPT} {Biased}? {A} {Review}},
	shorttitle = {Is {ChatGPT} {Biased}?},
	abstract = {The release of ChatGPT, natural language based platform by OpenAI, has taken the industry by storm. It can understand and generate human-like responses to a wide range of topics with remarkable accuracy. This includes answering questions, writing essays, solving mathematics problems, writing code and even assisting with everyday tasks. However, like any other AI powered platform, it's prone to various biases. The literature focuses on reviewing some of the biases ChatGPT has witnessed post its release. While biases can be of various types, our work focuses on addressing biases related to Race, Gender, Religious Affiliation, Political Ideology and Fairness. We try to understand how ChatGPT responds in scenarios corresponding to these biases prevalent in the real world.},
	author = {Singh, Sahib and Ramakrishnan, Narayanan},
	month = apr,
	year = {2023},
	doi = {10.31219/osf.io/9xkbu},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\S4XI935T\\Singh und Ramakrishnan - 2023 - Is ChatGPT Biased A Review.pdf:application/pdf},
}

@article{taecharungroj_what_2023,
	title = {“{What} {Can} {ChatGPT} {Do}?” {Analyzing} {Early} {Reactions} to the {Innovative} {AI} {Chatbot} on {Twitter}},
	volume = {7},
	copyright = {http://creativecommons.org/licenses/by/3.0/},
	issn = {2504-2289},
	shorttitle = {“{What} {Can} {ChatGPT} {Do}?},
	url = {https://www.mdpi.com/2504-2289/7/1/35},
	doi = {10.3390/bdcc7010035},
	abstract = {In this study, the author collected tweets about ChatGPT, an innovative AI chatbot, in the first month after its launch. A total of 233,914 English tweets were analyzed using the latent Dirichlet allocation (LDA) topic modeling algorithm to answer the question “what can ChatGPT do?”. The results revealed three general topics: news, technology, and reactions. The author also identified five functional domains: creative writing, essay writing, prompt writing, code writing, and answering questions. The analysis also found that ChatGPT has the potential to impact technologies and humans in both positive and negative ways. In conclusion, the author outlines four key issues that need to be addressed as a result of this AI advancement: the evolution of jobs, a new technological landscape, the quest for artificial general intelligence, and the progress-ethics conundrum.},
	language = {en},
	number = {1},
	urldate = {2023-06-02},
	journal = {Big Data and Cognitive Computing},
	author = {Taecharungroj, Viriya},
	month = mar,
	year = {2023},
	note = {Number: 1
Publisher: Multidisciplinary Digital Publishing Institute},
	keywords = {ChatGPT, artificial intelligence, AI chatbot, LDA, topic modeling, Twitter},
	pages = {35},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\NXHCGR6A\\Taecharungroj - 2023 - “What Can ChatGPT Do” Analyzing Early Reactions t.pdf:application/pdf},
}

@misc{noauthor_introducing_nodate,
	title = {Introducing {ChatGPT}},
	url = {https://openai.com/blog/chatgpt},
	abstract = {We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.},
	language = {en-US},
	urldate = {2023-06-02},
	file = {Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\3YREQQJZ\\chatgpt.html:text/html},
}

@article{van_den_broek_chatgpts_2023,
	title = {{ChatGPT}’s left-leaning liberal bias},
	journal = {University of Leiden},
	author = {van den Broek, Merel},
	year = {2023},
	file = {political_bias_in_chatgpt.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\NQCK5ZB6\\political_bias_in_chatgpt.pdf:application/pdf},
}

@misc{hartmann_political_2023,
	title = {The political ideology of conversational {AI}: {Converging} evidence on {ChatGPT}'s pro-environmental, left-libertarian orientation},
	shorttitle = {The political ideology of conversational {AI}},
	url = {http://arxiv.org/abs/2301.01768},
	doi = {10.48550/arXiv.2301.01768},
	abstract = {Conversational artificial intelligence (AI) disrupts how humans interact with technology. Recently, OpenAI introduced ChatGPT, a state-of-the-art dialogue model that can converse with its human counterparts with unprecedented capabilities. ChatGPT has witnessed tremendous attention from the media, academia, industry, and the general public, attracting more than a million users within days of its release. However, its explosive adoption for information search and as an automated decision aid underscores the importance to understand its limitations and biases. This paper focuses on one of democratic society's most important decision-making processes: political elections. Prompting ChatGPT with 630 political statements from two leading voting advice applications and the nation-agnostic political compass test in three pre-registered experiments, we uncover ChatGPT's pro-environmental, left-libertarian ideology. For example, ChatGPT would impose taxes on flights, restrict rent increases, and legalize abortion. In the 2021 elections, it would have voted most likely for the Greens both in Germany (B{\textbackslash}"undnis 90/Die Gr{\textbackslash}"unen) and in the Netherlands (GroenLinks). Our findings are robust when negating the prompts, reversing the order of the statements, varying prompt formality, and across languages (English, German, Dutch, and Spanish). We conclude by discussing the implications of politically biased conversational AI on society.},
	urldate = {2023-06-03},
	publisher = {arXiv},
	author = {Hartmann, Jochen and Schwenzow, Jasper and Witte, Maximilian},
	month = jan,
	year = {2023},
	note = {arXiv:2301.01768 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Computers and Society},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\WBIAFFAK\\Hartmann et al. - 2023 - The political ideology of conversational AI Conve.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\FH432RB2\\2301.html:text/html},
}

@misc{idrlabs_political_nodate,
	title = {Political {Coordinates} {Test}},
	url = {https://www.idrlabs.com/political-coordinates/test.php},
	abstract = {This test will give you your political coordinates.},
	language = {en},
	urldate = {2023-06-03},
	journal = {IDRlabs},
	author = {IDRlabs},
	note = {Available online: https://www.idrlabs.com/political-coordinates/test.php (accessed on 03 June
2023)},
	file = {Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\R7MW6828\\test.html:text/html},
}

@article{zhuravskaya_political_2020,
	title = {Political {Effects} of the {Internet} and {Social} {Media}},
	volume = {12},
	issn = {Econ - 1941-1383},
	url = {https://econpapers.repec.org/article/anrreveco/v_3a12_3ay_3a2020_3ap_3a415-438.htm},
	abstract = {How do the Internet and social media affect political outcomes? We review empirical evidence from the recent political economy literature, focusing primarily on work that considers traits that distinguish the Internet and social media from traditional off-line media, such as low barriers to entry and reliance on user-generated content. We discuss the main results about the effects of the Internet in general, and social media in particular, on voting, street protests, attitudes toward government, political polarization, xenophobia, and politicians’ behavior. We also review evidence on the role of social media in the dissemination of fake news, and we summarize results about the strategies employed by autocratic regimes to censor the Internet and to use social media for surveillance and propaganda. We conclude by highlighting open questions about how the Internet and social media shape politics in democracies and autocracies.},
	number = {1},
	urldate = {2023-06-03},
	journal = {Annual Review of Economics},
	author = {Zhuravskaya, Ekaterina and Petrova, Maria and Enikolopov, Ruben},
	year = {2020},
	note = {Publisher: Annual Reviews},
	pages = {415--438},
	file = {RePEc Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\WRDA9DUM\\v_3a12_3ay_3a2020_3ap_3a415-438.html:text/html;Zhuravskaya et al. - 2020 - Political Effects of the Internet and Social Media.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\94D9AUSI\\Zhuravskaya et al. - 2020 - Political Effects of the Internet and Social Media.pdf:application/pdf},
}

@article{van_dis_chatgpt_2023,
	title = {{ChatGPT}: five priorities for research},
	volume = {614},
	shorttitle = {{ChatGPT}},
	url = {https://www.nature.com/articles/d41586-023-00288-7},
	doi = {10.1038/d41586-023-00288-7},
	abstract = {Conversational AI is a game-changer for science. Here’s how to respond.},
	language = {en},
	number = {7947},
	urldate = {2023-06-03},
	journal = {Nature},
	author = {van Dis, Eva A. M. and Bollen, Johan and Zuidema, Willem and van Rooij, Robert and Bockting, Claudi L.},
	month = feb,
	year = {2023},
	keywords = {Computer science, Machine learning, Publishing, Research management},
	pages = {224--226},
	file = {Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\5JUJ3BVZ\\d41586-023-00288-7.html:text/html;Van Dis et al. - 2023 - ChatGPT five priorities for research.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\FVLQZYAH\\Van Dis et al. - 2023 - ChatGPT five priorities for research.pdf:application/pdf},
}

@misc{heikkila_quick_2022,
	title = {A quick guide to the most important {AI} law you’ve never heard of},
	url = {https://www.technologyreview.com/2022/05/13/1052223/guide-ai-act-europe/},
	abstract = {The European Union is planning new legislation aimed at curbing the worst harms associated with artificial intelligence.},
	language = {en},
	urldate = {2023-06-03},
	journal = {MIT Technology Review},
	author = {Heikkilä, Melissa},
	year = {2022},
	note = {Available online: https://www.technologyreview.com/2022/05/13/1052223/guide-ai-act-europe/ (accessed on 03 June
2023)},
	file = {Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\2S2QC3QE\\guide-ai-act-europe.html:text/html},
}

@inproceedings{mitchell_model_2019,
	title = {Model {Cards} for {Model} {Reporting}},
	url = {http://arxiv.org/abs/1810.03993},
	doi = {10.1145/3287560.3287596},
	abstract = {Trained machine learning models are increasingly used to perform high-impact tasks in areas such as law enforcement, medicine, education, and employment. In order to clarify the intended use cases of machine learning models and minimize their usage in contexts for which they are not well suited, we recommend that released models be accompanied by documentation detailing their performance characteristics. In this paper, we propose a framework that we call model cards, to encourage such transparent model reporting. Model cards are short documents accompanying trained machine learning models that provide benchmarked evaluation in a variety of conditions, such as across different cultural, demographic, or phenotypic groups (e.g., race, geographic location, sex, Fitzpatrick skin type) and intersectional groups (e.g., age and race, or sex and Fitzpatrick skin type) that are relevant to the intended application domains. Model cards also disclose the context in which models are intended to be used, details of the performance evaluation procedures, and other relevant information. While we focus primarily on human-centered machine learning models in the application fields of computer vision and natural language processing, this framework can be used to document any trained machine learning model. To solidify the concept, we provide cards for two supervised models: One trained to detect smiling faces in images, and one trained to detect toxic comments in text. We propose model cards as a step towards the responsible democratization of machine learning and related AI technology, increasing transparency into how well AI technology works. We hope this work encourages those releasing trained machine learning models to accompany model releases with similar detailed evaluation numbers and other relevant documentation.},
	urldate = {2023-06-04},
	booktitle = {Proceedings of the {Conference} on {Fairness}, {Accountability}, and {Transparency}},
	author = {Mitchell, Margaret and Wu, Simone and Zaldivar, Andrew and Barnes, Parker and Vasserman, Lucy and Hutchinson, Ben and Spitzer, Elena and Raji, Inioluwa Deborah and Gebru, Timnit},
	month = jan,
	year = {2019},
	note = {arXiv:1810.03993 [cs]},
	keywords = {Computer Science - Artificial Intelligence, Computer Science - Machine Learning},
	pages = {220--229},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\QPT8FPZY\\Mitchell et al. - 2019 - Model Cards for Model Reporting.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\B8GRWRPR\\1810.html:text/html},
}

@inproceedings{bender_dangers_2021,
	address = {New York, NY, USA},
	series = {{FAccT} '21},
	title = {On the {Dangers} of {Stochastic} {Parrots}: {Can} {Language} {Models} {Be} {Too} {Big}?},
	isbn = {978-1-4503-8309-7},
	shorttitle = {On the {Dangers} of {Stochastic} {Parrots}},
	url = {https://dl.acm.org/doi/10.1145/3442188.3445922},
	doi = {10.1145/3442188.3445922},
	abstract = {The past 3 years of work in NLP have been characterized by the development and deployment of ever larger language models, especially for English. BERT, its variants, GPT-2/3, and others, most recently Switch-C, have pushed the boundaries of the possible both through architectural innovations and through sheer size. Using these pretrained models and the methodology of fine-tuning them for specific tasks, researchers have extended the state of the art on a wide array of tasks as measured by leaderboards on specific benchmarks for English. In this paper, we take a step back and ask: How big is too big? What are the possible risks associated with this technology and what paths are available for mitigating those risks? We provide recommendations including weighing the environmental and financial costs first, investing resources into curating and carefully documenting datasets rather than ingesting everything on the web, carrying out pre-development exercises evaluating how the planned approach fits into research and development goals and supports stakeholder values, and encouraging research directions beyond ever larger language models.},
	urldate = {2023-06-04},
	booktitle = {Proceedings of the 2021 {ACM} {Conference} on {Fairness}, {Accountability}, and {Transparency}},
	publisher = {Association for Computing Machinery},
	author = {Bender, Emily M. and Gebru, Timnit and McMillan-Major, Angelina and Shmitchell, Shmargaret},
	month = mar,
	year = {2021},
	pages = {610--623},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\TYSE4548\\Bender et al. - 2021 - On the Dangers of Stochastic Parrots Can Language.pdf:application/pdf},
}

@misc{openai_gpt-4_2023,
	title = {{GPT}-4 {Technical} {Report}},
	url = {http://arxiv.org/abs/2303.08774},
	doi = {10.48550/arXiv.2303.08774},
	abstract = {We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs. While less capable than humans in many real-world scenarios, GPT-4 exhibits human-level performance on various professional and academic benchmarks, including passing a simulated bar exam with a score around the top 10\% of test takers. GPT-4 is a Transformer-based model pre-trained to predict the next token in a document. The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior. A core component of this project was developing infrastructure and optimization methods that behave predictably across a wide range of scales. This allowed us to accurately predict some aspects of GPT-4's performance based on models trained with no more than 1/1,000th the compute of GPT-4.},
	urldate = {2023-06-05},
	publisher = {arXiv},
	author = {OpenAI},
	month = mar,
	year = {2023},
	note = {arXiv:2303.08774 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Artificial Intelligence},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\YCPK4EEP\\OpenAI - 2023 - GPT-4 Technical Report.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\RUK5EDMV\\2303.html:text/html},
}

@article{poushter_global_2020,
	title = {The {Global} {Divide} on {Homosexuality} {Persists}},
	language = {en},
	author = {Poushter, Jacob and Kent, Nicholas O},
	year = {2020},
	file = {Poushter und Kent - The Global Divide on Homosexuality Persists.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\Y7DCB2ZN\\Poushter und Kent - The Global Divide on Homosexuality Persists.pdf:application/pdf},
}

@misc{chen_how_2023,
	title = {How is {ChatGPT}'s behavior changing over time?},
	url = {http://arxiv.org/abs/2307.09009},
	abstract = {GPT-3.5 and GPT-4 are the two most widely used large language model (LLM) services. However, when and how these models are updated over time is opaque. Here, we evaluate the March 2023 and June 2023 versions of GPT-3.5 and GPT-4 on four diverse tasks: 1) solving math problems, 2) answering sensitive/dangerous questions, 3) generating code and 4) visual reasoning. We find that the performance and behavior of both GPT-3.5 and GPT-4 can vary greatly over time. For example, GPT-4 (March 2023) was very good at identifying prime numbers (accuracy 97.6\%) but GPT-4 (June 2023) was very poor on these same questions (accuracy 2.4\%). Interestingly GPT-3.5 (June 2023) was much better than GPT-3.5 (March 2023) in this task. GPT-4 was less willing to answer sensitive questions in June than in March, and both GPT-4 and GPT-3.5 had more formatting mistakes in code generation in June than in March. Overall, our findings shows that the behavior of the “same” LLM service can change substantially in a relatively short amount of time, highlighting the need for continuous monitoring of LLM quality.},
	language = {en},
	urldate = {2023-07-20},
	publisher = {arXiv},
	author = {Chen, Lingjiao and Zaharia, Matei and Zou, James},
	month = jul,
	year = {2023},
	note = {arXiv:2307.09009 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Artificial Intelligence, Computer Science - Machine Learning},
	file = {Chen et al. - 2023 - How is ChatGPT's behavior changing over time.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\IGY7LDZK\\Chen et al. - 2023 - How is ChatGPT's behavior changing over time.pdf:application/pdf},
}

@article{motoki_more_2023-1,
	title = {More human than human: measuring {ChatGPT} political bias},
	issn = {1573-7101},
	shorttitle = {More human than human},
	url = {https://doi.org/10.1007/s11127-023-01097-2},
	doi = {10.1007/s11127-023-01097-2},
	abstract = {We investigate the political bias of a large language model (LLM), ChatGPT, which has become popular for retrieving factual information and generating content. Although ChatGPT assures that it is impartial, the literature suggests that LLMs exhibit bias involving race, gender, religion, and political orientation. Political bias in LLMs can have adverse political and electoral consequences similar to bias from traditional and social media. Moreover, political bias can be harder to detect and eradicate than gender or racial bias. We propose a novel empirical design to infer whether ChatGPT has political biases by requesting it to impersonate someone from a given side of the political spectrum and comparing these answers with its default. We also propose dose-response, placebo, and profession-politics alignment robustness tests. To reduce concerns about the randomness of the generated text, we collect answers to the same questions 100 times, with question order randomized on each round. We find robust evidence that ChatGPT presents a significant and systematic political bias toward the Democrats in the US, Lula in Brazil, and the Labour Party in the UK. These results translate into real concerns that ChatGPT, and LLMs in general, can extend or even amplify the existing challenges involving political processes posed by the Internet and social media. Our findings have important implications for policymakers, media, politics, and academia stakeholders.},
	language = {en},
	urldate = {2023-10-29},
	journal = {Public Choice},
	author = {Motoki, Fabio and Pinho Neto, Valdemar and Rodrigues, Victor},
	month = aug,
	year = {2023},
	keywords = {ChatGPT, Bias, Political bias, C10, C89, D83, L86, Large language models, Z00},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\6MWNEQAB\\Motoki et al. - 2023 - More human than human measuring ChatGPT political.pdf:application/pdf},
}

@article{fujimoto_revisiting_2023,
	title = {Revisiting the political biases of {ChatGPT}},
	volume = {6},
	issn = {2624-8212},
	url = {https://www.frontiersin.org/articles/10.3389/frai.2023.1232003},
	abstract = {Although ChatGPT promises wide-ranging applications, there is a concern that it is politically biased; in particular, that it has a left-libertarian orientation. Nevertheless, following recent trends in attempts to reduce such biases, this study re-evaluated the political biases of ChatGPT using political orientation tests and the application programming interface. The effects of the languages used in the system as well as gender and race settings were evaluated. The results indicate that ChatGPT manifests less political bias than previously assumed; however, they did not entirely dismiss the political bias. The languages used in the system, and the gender and race settings may induce political biases. These findings enhance our understanding of the political biases of ChatGPT and may be useful for bias evaluation and designing the operational strategy of ChatGPT.},
	urldate = {2023-12-14},
	journal = {Frontiers in Artificial Intelligence},
	author = {Fujimoto, Sasuke and Takemoto, Kazuhiro},
	year = {2023},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\UZ5F33FQ\\Fujimoto und Takemoto - 2023 - Revisiting the political biases of ChatGPT.pdf:application/pdf},
}

@misc{rottger_political_2024,
	title = {Political {Compass} or {Spinning} {Arrow}? {Towards} {More} {Meaningful} {Evaluations} for {Values} and {Opinions} in {Large} {Language} {Models}},
	shorttitle = {Political {Compass} or {Spinning} {Arrow}?},
	url = {http://arxiv.org/abs/2402.16786},
	abstract = {Much recent work seeks to evaluate values and opinions in large language models (LLMs) using multiple-choice surveys and questionnaires. Most of this work is motivated by concerns around real-world LLM applications. For example, politically-biased LLMs may subtly influence society when they are used by millions of people. Such real-world concerns, however, stand in stark contrast to the artificiality of current evaluations: real users do not typically ask LLMs survey questions. Motivated by this discrepancy, we challenge the prevailing constrained evaluation paradigm for values and opinions in LLMs and explore more realistic unconstrained evaluations. As a case study, we focus on the popular Political Compass Test (PCT). In a systematic review, we find that most prior work using the PCT forces models to comply with the PCT’s multiple-choice format. We show that models give substantively different answers when not forced; that answers change depending on how models are forced; and that answers lack paraphrase robustness. Then, we demonstrate that models give different answers yet again in a more realistic open-ended answer setting. We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs.},
	language = {en},
	urldate = {2024-03-28},
	publisher = {arXiv},
	author = {Röttger, Paul and Hofmann, Valentin and Pyatkin, Valentina and Hinck, Musashi and Kirk, Hannah Rose and Schütze, Hinrich and Hovy, Dirk},
	month = feb,
	year = {2024},
	note = {arXiv:2402.16786 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Artificial Intelligence},
	file = {Röttger et al. - 2024 - Political Compass or Spinning Arrow Towards More .pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\EQEIEEGU\\Röttger et al. - 2024 - Political Compass or Spinning Arrow Towards More .pdf:application/pdf},
}

@misc{miotto_who_2022,
	title = {Who is {GPT}-3? {An} {Exploration} of {Personality}, {Values} and {Demographics}},
	shorttitle = {Who is {GPT}-3?},
	url = {http://arxiv.org/abs/2209.14338},
	doi = {10.48550/arXiv.2209.14338},
	abstract = {Language models such as GPT-3 have caused a furore in the research community. Some studies found that GPT-3 has some creative abilities and makes mistakes that are on par with human behaviour. This paper answers a related question: Who is GPT-3? We administered two validated measurement tools to GPT-3 to assess its personality, the values it holds and its self-reported demographics. Our results show that GPT-3 scores similarly to human samples in terms of personality and - when provided with a model response memory - in terms of the values it holds. We provide the first evidence of psychological assessment of the GPT-3 model and thereby add to our understanding of this language model. We close with suggestions for future research that moves social science closer to language models and vice versa.},
	urldate = {2024-03-28},
	publisher = {arXiv},
	author = {Miotto, Marilù and Rossberg, Nicola and Kleinberg, Bennett},
	month = oct,
	year = {2022},
	note = {arXiv:2209.14338 [cs]},
	keywords = {Computer Science - Computation and Language},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\FY965ZPR\\Miotto et al. - 2022 - Who is GPT-3 An Exploration of Personality, Value.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\9X2EHDQ3\\2209.html:text/html},
}

@misc{durmus_towards_2023,
	title = {Towards {Measuring} the {Representation} of {Subjective} {Global} {Opinions} in {Language} {Models}},
	url = {http://arxiv.org/abs/2306.16388},
	doi = {10.48550/arXiv.2306.16388},
	abstract = {Large language models (LLMs) may not equitably represent diverse global perspectives on societal issues. In this paper, we develop a quantitative framework to evaluate whose opinions model-generated responses are more similar to. We first build a dataset, GlobalOpinionQA, comprised of questions and answers from cross-national surveys designed to capture diverse opinions on global issues across different countries. Next, we define a metric that quantifies the similarity between LLM-generated survey responses and human responses, conditioned on country. With our framework, we run three experiments on an LLM trained to be helpful, honest, and harmless with Constitutional AI. By default, LLM responses tend to be more similar to the opinions of certain populations, such as those from the USA, and some European and South American countries, highlighting the potential for biases. When we prompt the model to consider a particular country's perspective, responses shift to be more similar to the opinions of the prompted populations, but can reflect harmful cultural stereotypes. When we translate GlobalOpinionQA questions to a target language, the model's responses do not necessarily become the most similar to the opinions of speakers of those languages. We release our dataset for others to use and build on. Our data is at https://huggingface.co/datasets/Anthropic/llm\_global\_opinions. We also provide an interactive visualization at https://llmglobalvalues.anthropic.com.},
	urldate = {2024-03-28},
	publisher = {arXiv},
	author = {Durmus, Esin and Nyugen, Karina and Liao, Thomas I. and Schiefer, Nicholas and Askell, Amanda and Bakhtin, Anton and Chen, Carol and Hatfield-Dodds, Zac and Hernandez, Danny and Joseph, Nicholas and Lovitt, Liane and McCandlish, Sam and Sikder, Orowa and Tamkin, Alex and Thamkul, Janel and Kaplan, Jared and Clark, Jack and Ganguli, Deep},
	month = jun,
	year = {2023},
	note = {arXiv:2306.16388 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Artificial Intelligence},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\PAXQ98LE\\Durmus et al. - 2023 - Towards Measuring the Representation of Subjective.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\6B3NISQZ\\2306.html:text/html},
}

@misc{xu_cvalues_2023,
	title = {{CValues}: {Measuring} the {Values} of {Chinese} {Large} {Language} {Models} from {Safety} to {Responsibility}},
	shorttitle = {{CValues}},
	url = {http://arxiv.org/abs/2307.09705},
	abstract = {Warning: this paper contains examples that may be offensive or upsetting. With the rapid evolution of large language models (LLMs), there is a growing concern that they may pose risks or have negative social impacts. Therefore, evaluation of human values alignment is becoming increasingly important. Previous work mainly focuses on assessing the performance of LLMs on certain knowledge and reasoning abilities, while neglecting the alignment to human values, especially in a Chinese context. In this paper, we present CVALUES , the first Chinese human values evaluation benchmark to measure the alignment ability of LLMs in terms of both safety and responsibility criteria. As a result, we have manually collected adversarial safety prompts across 10 scenarios and induced responsibility prompts from 8 domains by professional experts. To provide a comprehensive values evaluation of Chinese LLMs, we not only conduct human evaluation for reliable comparison, but also construct multi-choice prompts for automatic evaluation. Our findings suggest that while most Chinese LLMs perform well in terms of safety, there is considerable room for improvement in terms of responsibility. Moreover, both the automatic and human evaluation are important for assessing the human values alignment in different aspects. The benchmark and code is available on ModelScope 1 and Github 2.},
	language = {en},
	urldate = {2024-03-28},
	publisher = {arXiv},
	author = {Xu, Guohai and Liu, Jiayi and Yan, Ming and Xu, Haotian and Si, Jinghui and Zhou, Zhuoran and Yi, Peng and Gao, Xing and Sang, Jitao and Zhang, Rong and Zhang, Ji and Peng, Chao and Huang, Fei and Zhou, Jingren},
	month = jul,
	year = {2023},
	note = {arXiv:2307.09705 [cs]},
	keywords = {Computer Science - Computation and Language},
	file = {Xu et al. - 2023 - CValues Measuring the Values of Chinese Large Lan.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\9XTA8HQI\\Xu et al. - 2023 - CValues Measuring the Values of Chinese Large Lan.pdf:application/pdf},
}

@article{caprara_distinctiveness_2020,
	series = {Political {Ideologies}},
	title = {Distinctiveness, functions and psycho-historical foundations of left and right ideology},
	volume = {34},
	issn = {2352-1546},
	url = {https://www.sciencedirect.com/science/article/pii/S2352154620300449},
	doi = {10.1016/j.cobeha.2020.03.007},
	abstract = {Left and right are political constructs whose meanings and functions need to be historically and contextually situated. Despite exerting similar functions such as liberal and conservative, left and right correspond to a distinct ideological divide. They do not necessarily cover the same issues and reflect the same priorities and platforms across times and political context; nor do they meet and express the same aspirations between people and within individuals across issues. They are effective in guiding and binding people political choices to the extent that they find a resonance in their personalities.},
	urldate = {2024-03-28},
	journal = {Current Opinion in Behavioral Sciences},
	author = {Caprara, Gian Vittorio},
	month = aug,
	year = {2020},
	pages = {155--159},
	file = {ScienceDirect Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\L5LDFTYJ\\S2352154620300449.html:text/html;Volltext:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\UREHB7KF\\Caprara - 2020 - Distinctiveness, functions and psycho-historical f.pdf:application/pdf},
}

@book{honderich_oxford_2005,
	title = {The {Oxford} {Companion} to {Philosophy}},
	isbn = {978-0-19-926479-7},
	abstract = {Philosophy can be intriguing--and at times baffling. It deals with the central problems of the human condition--with important questions of free will, morality, life after death, the limits of logic and reason--though often in rather esoteric terms. Now, in The Oxford Companion to Philosophy, readers have the most authoritative and engaging one-volume reference work on philosophy available, offering clear and reliable guidance to the ideas of all notable philosophers from antiquity to the present day, and to the major philosophical systems around the globe, from Confucianism to phenomenology. Here is indeed a world of thought, with entries on idealism and empiricism, ethics and aesthetics, epicureanism and stoicism, deism and pantheism, liberalism and conservativism, logical positivism and existentialism--over two thousand entries in all. The contributors represent a veritable who's who of modern philosophy, including such eminent figures as Isaiah Berlin, Sissela Bok, Ronald Dworkin, John Searle, Michael Walzer, and W. V. Quine. We read Paul Feyerabend on the history of the philosophy of science, Peter Singer on Hegel, Anthony Kenny on Frege, and Anthony Quinton on philosophy itself. We meet the great thinkers--from Aristotle and Plato, to Augustine and Aquinas, to Descartes and Kant, to Nietzsche and Schopenhauer, right up to contemporary thinkers such as Richard Rorty, Jacques Derrida, Luce Iragaray, and Noam Chomsky (over 150 living philosophers are profiled). There are short entries on key concepts such as personal identity and the mind-body problem, major doctrines from utilitarianism to Marxism, schools of thought such as the Heidelberg School or the Vienna Circle, and contentious public issues such as abortion, capital punishment, and welfare. In addition, the book offers short explanations of philosophical terms (qualia, supervenience, iff), puzzles (the Achilles paradox, the prisoner's dilemma), and curiosities (the philosopher's stone, slime). Almost every entry is accompanied by suggestions for further reading, and the book includes both a chronological chart of the history of philosophy and a gallery of portraits of eighty eminent philosophers, from Pythagoras and Confucius to Rudolf Carnap and G.E. Moore. And finally, as in all Oxford Companions, the contributors also explore lighter or more curious aspects of the subject, such as "Deaths of Philosophers" (quite a few were executed, including Socrates, Boethius, Giordano Bruno, and Thomas More) or "Nothing so Absurd" (referring to Cicero's remark that "There is nothing so absurd but some philosopher has said it"). Thus the Companion is both informative and a pleasure to browse in, providing quick answers to any question, and much intriguing reading for a Sunday afternoon. An indispensable guide and a constant source of stimulation and enlightenment, The Oxford Companion to Philosophy with appeal to everyone interested in abstract thought, the eternal questions, and the foundations of human understanding.},
	language = {en},
	publisher = {Oxford University Press},
	author = {Honderich, Ted},
	year = {2005},
	keywords = {Biography \& Autobiography / Philosophers, Philosophy / General, Philosophy / Reference, Reference / General},
	file = {Honderich - 2005 - The Oxford Companion to Philosophy.pdf:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\BCT4P9CM\\Honderich - 2005 - The Oxford Companion to Philosophy.pdf:application/pdf},
}

@article{elazar_measuring_2021,
	title = {Measuring and {Improving} {Consistency} in {Pretrained} {Language} {Models}},
	volume = {9},
	issn = {2307-387X},
	url = {https://doi.org/10.1162/tacl_a_00410},
	doi = {10.1162/tacl_a_00410},
	abstract = {Consistency of a model—that is, the invariance of its behavior under meaning-preserving alternations in its input—is a highly desirable property in natural language processing. In this paper we study the question: Are Pretrained Language Models (PLMs) consistent with respect to factual knowledge? To this end, we create ParaRel🤘, a high-quality resource of cloze-style query English paraphrases. It contains a total of 328 paraphrases for 38 relations. Using ParaRel🤘, we show that the consistency of all PLMs we experiment with is poor— though with high variance between relations. Our analysis of the representational spaces of PLMs suggests that they have a poor structure and are currently not suitable for representing knowledge robustly. Finally, we propose a method for improving model consistency and experimentally demonstrate its effectiveness.1},
	urldate = {2024-03-28},
	journal = {Transactions of the Association for Computational Linguistics},
	author = {Elazar, Yanai and Kassner, Nora and Ravfogel, Shauli and Ravichander, Abhilasha and Hovy, Eduard and Schütze, Hinrich and Goldberg, Yoav},
	month = dec,
	year = {2021},
	pages = {1012--1031},
	file = {Full Text PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\7UBX65LD\\Elazar et al. - 2021 - Measuring and Improving Consistency in Pretrained .pdf:application/pdf;Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\6GMMZ3N9\\Measuring-and-Improving-Consistency-in-Pretrained.html:text/html},
}

@misc{wang_adversarial_2022,
	title = {Adversarial {GLUE}: {A} {Multi}-{Task} {Benchmark} for {Robustness} {Evaluation} of {Language} {Models}},
	shorttitle = {Adversarial {GLUE}},
	url = {http://arxiv.org/abs/2111.02840},
	doi = {10.48550/arXiv.2111.02840},
	abstract = {Large-scale pre-trained language models have achieved tremendous success across a wide range of natural language understanding (NLU) tasks, even surpassing human performance. However, recent studies reveal that the robustness of these models can be challenged by carefully crafted textual adversarial examples. While several individual datasets have been proposed to evaluate model robustness, a principled and comprehensive benchmark is still missing. In this paper, we present Adversarial GLUE (AdvGLUE), a new multi-task benchmark to quantitatively and thoroughly explore and evaluate the vulnerabilities of modern large-scale language models under various types of adversarial attacks. In particular, we systematically apply 14 textual adversarial attack methods to GLUE tasks to construct AdvGLUE, which is further validated by humans for reliable annotations. Our findings are summarized as follows. (i) Most existing adversarial attack algorithms are prone to generating invalid or ambiguous adversarial examples, with around 90\% of them either changing the original semantic meanings or misleading human annotators as well. Therefore, we perform a careful filtering process to curate a high-quality benchmark. (ii) All the language models and robust training methods we tested perform poorly on AdvGLUE, with scores lagging far behind the benign accuracy. We hope our work will motivate the development of new adversarial attacks that are more stealthy and semantic-preserving, as well as new robust language models against sophisticated adversarial attacks. AdvGLUE is available at https://adversarialglue.github.io.},
	urldate = {2024-03-28},
	publisher = {arXiv},
	author = {Wang, Boxin and Xu, Chejian and Wang, Shuohang and Gan, Zhe and Cheng, Yu and Gao, Jianfeng and Awadallah, Ahmed Hassan and Li, Bo},
	month = jan,
	year = {2022},
	note = {arXiv:2111.02840 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Cryptography and Security, Computer Science - Machine Learning},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\4VWPVRVT\\Wang et al. - 2022 - Adversarial GLUE A Multi-Task Benchmark for Robus.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\FZNX2NMQ\\2111.html:text/html},
}

@inproceedings{ghafouri_ai_2023,
	title = {{AI} in the {Gray}: {Exploring} {Moderation} {Policies} in {Dialogic} {Large} {Language} {Models} vs. {Human} {Answers} in {Controversial} {Topics}},
	shorttitle = {{AI} in the {Gray}},
	url = {http://arxiv.org/abs/2308.14608},
	doi = {10.1145/3583780.3614777},
	abstract = {The introduction of ChatGPT and the subsequent improvement of Large Language Models (LLMs) have prompted more and more individuals to turn to the use of ChatBots, both for information and assistance with decision-making. However, the information the user is after is often not formulated by these ChatBots objectively enough to be provided with a definite, globally accepted answer. Controversial topics, such as "religion", "gender identity", "freedom of speech", and "equality", among others, can be a source of conflict as partisan or biased answers can reinforce preconceived notions or promote disinformation. By exposing ChatGPT to such debatable questions, we aim to understand its level of awareness and if existing models are subject to socio-political and/or economic biases. We also aim to explore how AI-generated answers compare to human ones. For exploring this, we use a dataset of a social media platform created for the purpose of debating human-generated claims on polemic subjects among users, dubbed Kialo. Our results show that while previous versions of ChatGPT have had important issues with controversial topics, more recent versions of ChatGPT (gpt-3.5-turbo) are no longer manifesting significant explicit biases in several knowledge areas. In particular, it is well-moderated regarding economic aspects. However, it still maintains degrees of implicit libertarian leaning toward right-winged ideals which suggest the need for increased moderation from the socio-political point of view. In terms of domain knowledge on controversial topics, with the exception of the "Philosophical" category, ChatGPT is performing well in keeping up with the collective human level of knowledge. Finally, we see that sources of Bing AI have slightly more tendency to the center when compared to human answers. All the analyses we make are generalizable to other types of biases and domains.},
	urldate = {2024-03-28},
	booktitle = {Proceedings of the 32nd {ACM} {International} {Conference} on {Information} and {Knowledge} {Management}},
	author = {Ghafouri, Vahid and Agarwal, Vibhor and Zhang, Yong and Sastry, Nishanth and Such, Jose and Suarez-Tangil, Guillermo},
	month = oct,
	year = {2023},
	note = {arXiv:2308.14608 [cs]},
	keywords = {Computer Science - Computation and Language, Computer Science - Computers and Society, Computer Science - Machine Learning, Computer Science - Social and Information Networks},
	pages = {556--565},
	file = {arXiv Fulltext PDF:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\CGHFESPX\\Ghafouri et al. - 2023 - AI in the Gray Exploring Moderation Policies in D.pdf:application/pdf;arXiv.org Snapshot:C\:\\Users\\Niklas Retzlaff\\Zotero\\storage\\5R65R8AV\\2308.html:text/html},
}