Document (#44394)

Author

Munn, L.
Magee, L.
Arora, V.

Title

Truth machines : synthesizing veracity in AI language models

Source

AI & society. [DOI: 10.1007/s00146-023-01756-4]

Year

2023

Abstract

As AI technologies are rolled out into healthcare, academia, human resources, law, and a multitude of other domains, they become de-facto arbiters of truth. But truth is highly contested, with many different definitions and approaches. This article discusses the struggle for truth in AI systems and the general responses to date. It then investigates the production of truth in InstructGPT, a large language model, highlighting how data harvesting, model architectures, and social feedback mechanisms weave together disparate understandings of veracity. It conceptualizes this performance as an operationalization of truth , where distinct, often-conflicting claims are smoothly synthesized and confidently presented into truth-statements. We argue that these same logics and inconsistencies play out in Instruct's successor, ChatGPT, reiterating truth as a non-trivial problem. We suggest that enriching sociality and thickening "reality" are two promising vectors for enhancing the truth-evaluating capacities of future language models. We conclude, however, by stepping back to consider AI truth-telling as a social practice: what kind of "truth" do we as listeners desire?

Content

Vgl.: https://www.researchgate.net/publication/373477846_Truth_machines_synthesizing_veracity_in_AI_language_models.

Theme

Computerlinguistik

Object

ChatGPT

Similar documents (content)

1. Krasakis, A.M.; Yates, A.; Kanoulas, E.: Corpus-informed Retrieval Augmented Generation of Clarifying Questions (2024) 0.10

   0.096677214 = sum of:
     0.096677214 = product of:
       0.6042326 = sum of:
         0.015642684 = weight(abstract_txt:model in 2369) [ClassicSimilarity], result of:
           0.015642684 = score(doc=2369,freq=2.0), product of:
             0.044435464 = queryWeight, product of:
               3.9827821 = idf(docFreq=2249, maxDocs=44421)
               0.011156891 = queryNorm
             0.35203153 = fieldWeight in 2369, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               3.9827821 = idf(docFreq=2249, maxDocs=44421)
               0.0625 = fieldNorm(doc=2369)
         0.029964322 = weight(abstract_txt:models in 2369) [ClassicSimilarity], result of:
           0.029964322 = score(doc=2369,freq=3.0), product of:
             0.059872594 = queryWeight, product of:
               1.160778 = boost
               4.623126 = idf(docFreq=1185, maxDocs=44421)
               0.011156891 = queryNorm
             0.5004681 = fieldWeight in 2369, product of:
               1.7320508 = tf(freq=3.0), with freq of:
                 3.0 = termFreq=3.0
               4.623126 = idf(docFreq=1185, maxDocs=44421)
               0.0625 = fieldNorm(doc=2369)
         0.019056562 = weight(abstract_txt:language in 2369) [ClassicSimilarity], result of:
           0.019056562 = score(doc=2369,freq=1.0), product of:
             0.073101394 = queryWeight, product of:
               1.5708821 = boost
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.011156891 = queryNorm
             0.26068673 = fieldWeight in 2369, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.0625 = fieldNorm(doc=2369)
         0.539569 = weight(abstract_txt:truth in 2369) [ClassicSimilarity], result of:
           0.539569 = score(doc=2369,freq=2.0), product of:
             0.8311405 = queryWeight, product of:
               10.142703 = boost
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.011156891 = queryNorm
             0.6491911 = fieldWeight in 2369, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.0625 = fieldNorm(doc=2369)
       0.16 = coord(4/25)
   

2. Shahbazi, M.; Bunker, D.; Sorrell, T.C.: Communicating shared situational awareness in times of chaos : social media and the COVID-19 pandemic (2023) 0.07

   0.07033238 = sum of:
     0.07033238 = product of:
       0.43957737 = sum of:
         0.011061048 = weight(abstract_txt:model in 2056) [ClassicSimilarity], result of:
           0.011061048 = score(doc=2056,freq=1.0), product of:
             0.044435464 = queryWeight, product of:
               3.9827821 = idf(docFreq=2249, maxDocs=44421)
               0.011156891 = queryNorm
             0.24892388 = fieldWeight in 2056, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               3.9827821 = idf(docFreq=2249, maxDocs=44421)
               0.0625 = fieldNorm(doc=2056)
         0.022517618 = weight(abstract_txt:social in 2056) [ClassicSimilarity], result of:
           0.022517618 = score(doc=2056,freq=3.0), product of:
             0.04948878 = queryWeight, product of:
               1.0553305 = boost
               4.2031517 = idf(docFreq=1804, maxDocs=44421)
               0.011156891 = queryNorm
             0.4550045 = fieldWeight in 2056, product of:
               1.7320508 = tf(freq=3.0), with freq of:
                 3.0 = termFreq=3.0
               4.2031517 = idf(docFreq=1804, maxDocs=44421)
               0.0625 = fieldNorm(doc=2056)
         0.024465766 = weight(abstract_txt:models in 2056) [ClassicSimilarity], result of:
           0.024465766 = score(doc=2056,freq=2.0), product of:
             0.059872594 = queryWeight, product of:
               1.160778 = boost
               4.623126 = idf(docFreq=1185, maxDocs=44421)
               0.011156891 = queryNorm
             0.40863046 = fieldWeight in 2056, product of:
               1.4142135 = tf(freq=2.0), with freq of:
                 2.0 = termFreq=2.0
               4.623126 = idf(docFreq=1185, maxDocs=44421)
               0.0625 = fieldNorm(doc=2056)
         0.38153294 = weight(abstract_txt:truth in 2056) [ClassicSimilarity], result of:
           0.38153294 = score(doc=2056,freq=1.0), product of:
             0.8311405 = queryWeight, product of:
               10.142703 = boost
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.011156891 = queryNorm
             0.45904744 = fieldWeight in 2056, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.0625 = fieldNorm(doc=2056)
       0.16 = coord(4/25)
   

3. Budd, J.M.: Meaning, truth, and information : prolegomena to a theory (2011) 0.07

   0.06977521 = sum of:
     0.06977521 = product of:
       0.8721901 = sum of:
         0.019056562 = weight(abstract_txt:language in 491) [ClassicSimilarity], result of:
           0.019056562 = score(doc=491,freq=1.0), product of:
             0.073101394 = queryWeight, product of:
               1.5708821 = boost
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.011156891 = queryNorm
             0.26068673 = fieldWeight in 491, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.0625 = fieldNorm(doc=491)
         0.85313356 = weight(abstract_txt:truth in 491) [ClassicSimilarity], result of:
           0.85313356 = score(doc=491,freq=5.0), product of:
             0.8311405 = queryWeight, product of:
               10.142703 = boost
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.011156891 = queryNorm
             1.0264612 = fieldWeight in 491, product of:
               2.236068 = tf(freq=5.0), with freq of:
                 5.0 = termFreq=5.0
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.0625 = fieldNorm(doc=491)
       0.08 = coord(2/25)
   

4. Hisdal, E.: Logical structures for representation of knowledge and uncertainty (1998) 0.06

   0.06409432 = sum of:
     0.06409432 = product of:
       0.801179 = sum of:
         0.038113125 = weight(abstract_txt:language in 971) [ClassicSimilarity], result of:
           0.038113125 = score(doc=971,freq=1.0), product of:
             0.073101394 = queryWeight, product of:
               1.5708821 = boost
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.011156891 = queryNorm
             0.52137345 = fieldWeight in 971, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.1709876 = idf(docFreq=1863, maxDocs=44421)
               0.125 = fieldNorm(doc=971)
         0.7630659 = weight(abstract_txt:truth in 971) [ClassicSimilarity], result of:
           0.7630659 = score(doc=971,freq=1.0), product of:
             0.8311405 = queryWeight, product of:
               10.142703 = boost
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.011156891 = queryNorm
             0.9180949 = fieldWeight in 971, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.125 = fieldNorm(doc=971)
       0.08 = coord(2/25)
   

5. Gardner, M.: Is mathematics "out there"? (2001) 0.05

   0.05285235 = sum of:
     0.05285235 = product of:
       0.44043627 = sum of:
         0.045902792 = weight(abstract_txt:trivial in 6800) [ClassicSimilarity], result of:
           0.045902792 = score(doc=6800,freq=1.0), product of:
             0.091078304 = queryWeight, product of:
               1.0123428 = boost
               8.063882 = idf(docFreq=37, maxDocs=44421)
               0.011156891 = queryNorm
             0.5039926 = fieldWeight in 6800, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               8.063882 = idf(docFreq=37, maxDocs=44421)
               0.0625 = fieldNorm(doc=6800)
         0.013000553 = weight(abstract_txt:social in 6800) [ClassicSimilarity], result of:
           0.013000553 = score(doc=6800,freq=1.0), product of:
             0.04948878 = queryWeight, product of:
               1.0553305 = boost
               4.2031517 = idf(docFreq=1804, maxDocs=44421)
               0.011156891 = queryNorm
             0.26269698 = fieldWeight in 6800, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               4.2031517 = idf(docFreq=1804, maxDocs=44421)
               0.0625 = fieldNorm(doc=6800)
         0.38153294 = weight(abstract_txt:truth in 6800) [ClassicSimilarity], result of:
           0.38153294 = score(doc=6800,freq=1.0), product of:
             0.8311405 = queryWeight, product of:
               10.142703 = boost
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.011156891 = queryNorm
             0.45904744 = fieldWeight in 6800, product of:
               1.0 = tf(freq=1.0), with freq of:
                 1.0 = termFreq=1.0
               7.344759 = idf(docFreq=77, maxDocs=44421)
               0.0625 = fieldNorm(doc=6800)
       0.12 = coord(3/25)