Anders Søgaard
Biographic Data
| ID | 5995175 |
|---|---|
| NAME | Anders Søgaard |
| GIVEN NAMES | Anders |
| FAMILY NAME | Søgaard |
| SIGNATURE | SØGAARD A |
| AFFILIATIONS | University of Copenhagen |
| ORCID | 0000-0001-5250-4276 |
| VERIFIED | Yes |
| TOTAL WORKS | 9 |
| TOTAL CITATIONS | 1 |
| AUTHOR COUNT | 9 |
| EDITOR COUNT | 0 |
| FIRST PUBLICATION YEAR | 2005 |
| LATEST PUBLICATION YEAR | 2025 |
| H-INDEX | 1 |
On Hedden's proof that machine learning fairness metrics are flawed
Brian Hedden, in a recent article in Philosophy and Public Affairs [Hedden Citation2021. “On Statistical Criteria of Algorithmic Fairness.” Philosophy and Public Affairs 49 (2): 209–231. https://doi.org/10.1111/papa.v49.2.], presented a thought experiment designed to probe the validity of the fairness metrics used in machine learning (ML). The thought experiment has caused a great stir, also within machine learning [Viganó et al. “People are Not …
Externalist XAI
Developers of artificial intelligence (AI) often cannot explain the inferences their neural networks make, at least not in ways that satisfy user needs. XAI—explainable artificial intelligence—aims to develop techniques for providing such explanations. XAI researchers have adopted techniques that, to philosophers, seem representationalist–internalist, leading some philosophers to call for more externalist alternatives. But is explaining AI models…
Language acquisition in vector space
Language models are mathematical functions and, as such, induce vector spaces in which input is embedded. Comparing the point clouds of concept vectors across such language models and similar computer vision models, we see surprising similarities. This sheds new light on the Innateness Debate. Much linguistic structure can be induced from extra-linguistic data. Language models are generally thought to be too sample-inefficient to be good models o…
What is lost when a language dies
Nowak argues that the problem with language loss is not linguists’ loss of data or that the loss of a language is often a result of discrimination against its speakers. Instead, the real problem is its speakers’ loss of illocutionary force. I argue that Nowak’s argument rests on two premises that are both empirically unjustified: that cultural knowledge is a prerequisite for illocutionary force, and that language is a prerequisite for illocutiona…
On the Opacity of Deep Neural Networks
Deep neural networks are said to be opaque, impeding the development of safe and trustworthy artificial intelligence, but where this opacity stems from is less clear. What are the sufficient properties for neural network opacity? Here, I discuss five common properties of deep neural networks and two different kinds of opacity. Which of these properties are sufficient for what type of opacity? I show how each kind of opacity stems from only one of…
Understanding models understanding language
Landgrebe and Smith (Synthese 198(March):2061–2081, 2021) present an unflattering diagnosis of recent advances in what they call language-centric artificial intelligence—perhaps more widely known as natural language processing: The models that are currently employed do not have sufficient expressivity, will not generalize, and are fundamentally unable to induce linguistic semantics, they say. The diagnosis is mainly derived from an analysis of th…
Compound constructions: A reply to Bundgaard et al
In an article recently published in this journal, Bundgaard et al. present a series of arguments against non-compositional theories of (noun-noun) compound semantics. In consequence, they present a compositional alternative, where compositionality is relativized a bit to also take constructional semantics into account. I am fully convinced this is the right way to go about it, but unfortunately, the authors remain unclear about the exact nature o…
A tractable typed feature structure grammar for Mainland Scandinavian
This article presents a typed feature structure grammar formalism for Mainland Scandinavian which outperforms existing grammars in several respects: it is multilingual and captures dialectal variation and many typological facts, and it is computationally efficient. Our reference point is the grammar formalism of Underwood (1997). It is proven that Underwood's formalism is intractable. Our formalism improves on this result, i.e. it is decidable in…
The Semantics of Possession in Natural Language and Knowledge Representation
The semantics of possession is explored in a representative set of languages to evaluate two contemporary theories. It is evidenced that the theories describe different aspects of the universal possessive system, and that both provide relevant theoretical constructs for the vocabulary of linguistic typology. Our findings also have consequences for knowledge representation, and in the appendix, our new vocabulary is employed in the analysis of rel…
Understanding models understanding language
Landgrebe and Smith (Synthese 198(March):2061–2081, 2021) present an unflattering diagnosis of recent advances in what they call language-centric artificial intelligence—perhaps more widely known as natural language processing: The models that are currently employed do not have sufficient expressivity, will not generalize, and are fundamentally unable to induce linguistic semantics, they say. The diagnosis is mainly derived from an analysis of th…
The Semantics of Possession in Natural Language and Knowledge Representation
The semantics of possession is explored in a representative set of languages to evaluate two contemporary theories. It is evidenced that the theories describe different aspects of the universal possessive system, and that both provide relevant theoretical constructs for the vocabulary of linguistic typology. Our findings also have consequences for knowledge representation, and in the appendix, our new vocabulary is employed in the analysis of rel…
A tractable typed feature structure grammar for Mainland Scandinavian
This article presents a typed feature structure grammar formalism for Mainland Scandinavian which outperforms existing grammars in several respects: it is multilingual and captures dialectal variation and many typological facts, and it is computationally efficient. Our reference point is the grammar formalism of Underwood (1997). It is proven that Underwood's formalism is intractable. Our formalism improves on this result, i.e. it is decidable in…
Compound constructions: A reply to Bundgaard et al
In an article recently published in this journal, Bundgaard et al. present a series of arguments against non-compositional theories of (noun-noun) compound semantics. In consequence, they present a compositional alternative, where compositionality is relativized a bit to also take constructional semantics into account. I am fully convinced this is the right way to go about it, but unfortunately, the authors remain unclear about the exact nature o…
Understanding models understanding language
Landgrebe and Smith (Synthese 198(March):2061–2081, 2021) present an unflattering diagnosis of recent advances in what they call language-centric artificial intelligence—perhaps more widely known as natural language processing: The models that are currently employed do not have sufficient expressivity, will not generalize, and are fundamentally unable to induce linguistic semantics, they say. The diagnosis is mainly derived from an analysis of th…
On the Opacity of Deep Neural Networks
Deep neural networks are said to be opaque, impeding the development of safe and trustworthy artificial intelligence, but where this opacity stems from is less clear. What are the sufficient properties for neural network opacity? Here, I discuss five common properties of deep neural networks and two different kinds of opacity. Which of these properties are sufficient for what type of opacity? I show how each kind of opacity stems from only one of…
What is lost when a language dies
Nowak argues that the problem with language loss is not linguists’ loss of data or that the loss of a language is often a result of discrimination against its speakers. Instead, the real problem is its speakers’ loss of illocutionary force. I argue that Nowak’s argument rests on two premises that are both empirically unjustified: that cultural knowledge is a prerequisite for illocutionary force, and that language is a prerequisite for illocutiona…
On Hedden's proof that machine learning fairness metrics are flawed
Brian Hedden, in a recent article in Philosophy and Public Affairs [Hedden Citation2021. “On Statistical Criteria of Algorithmic Fairness.” Philosophy and Public Affairs 49 (2): 209–231. https://doi.org/10.1111/papa.v49.2.], presented a thought experiment designed to probe the validity of the fairness metrics used in machine learning (ML). The thought experiment has caused a great stir, also within machine learning [Viganó et al. “People are Not …
Externalist XAI
Developers of artificial intelligence (AI) often cannot explain the inferences their neural networks make, at least not in ways that satisfy user needs. XAI—explainable artificial intelligence—aims to develop techniques for providing such explanations. XAI researchers have adopted techniques that, to philosophers, seem representationalist–internalist, leading some philosophers to call for more externalist alternatives. But is explaining AI models…
Language acquisition in vector space
Language models are mathematical functions and, as such, induce vector spaces in which input is embedded. Comparing the point clouds of concept vectors across such language models and similar computer vision models, we see surprising similarities. This sheds new light on the Innateness Debate. Much linguistic structure can be induced from extra-linguistic data. Language models are generally thought to be too sample-inefficient to be good models o…
Computer Science (7 works) · Philosophy (7 works) · Linguistics (6 works) · Natural Language Processing Techniques (5 works) · Psychology (5 works) · Artificial Intelligence (4 works) · Epistemology (3 works) · Explainable Artificial Intelligence (XAI (3 works) · Artificial Intelligence (2 works) · Natural language processing (2 works)